test(kms): move the Vault KV2 Transit-wrapping doc guard into check_fips_wording.sh

`test_vault_kv2_sources_do_not_claim_transit_wrapping` asserted that four `include_str!`-pinned files never describe the Vault KV2 backend as wrapping key material through Vault's Transit engine. The invariant is a documentation-claim invariant with no behavioral twin by construction, and the test form was weak in both directions: it saw only four files (the same prose in a fifth file passed silently) and it stopped compiling — rather than reporting a violation — as soon as one of them was renamed. Move the four literals verbatim into `scripts/check_fips_wording.sh`, which already guards the adjacent cryptographic over-claim class (unsupported FIPS validation wording) and is anchored to the same policy document. The guard now greps every file under `crates/kms` for the same four case-sensitive literals and separately reports a moved pinned source instead of failing to build. `check_fips_wording.sh` previously ran only in `make pre-commit` / `pre-pr`, so wire it into the Quick Checks job of both CI workflows to keep the invariant's failure visibility at least as strong as the deleted test's.
test(e2e): drop 187 no-op serial markers from the three densest e2e suites (#6209 )
2026-08-18 10:43:15 +00:00 · 2026-08-18 17:30:11 +08:00 · 2026-08-18 17:05:30 +08:00 · 2026-08-18 17:04:55 +08:00 · 2026-08-18 08:14:05 +00:00 · 2026-08-18 16:09:30 +08:00
255 changed files with 15109 additions and 6998 deletions
@@ -66,8 +66,8 @@ s3s-footprint-check: ## Check the s3s dependency footprint ratchet stays frozen
 	./scripts/check_s3s_footprint.sh

 .PHONY: fips-wording-check
-fips-wording-check: ## Check outward docs do not make unsupported FIPS claims
-	@echo "📣 Checking FIPS wording guard..."
+fips-wording-check: ## Check docs and crates/kms do not over-claim crypto capabilities
+	@echo "📣 Checking cryptographic capability wording guard..."
 	./scripts/check_fips_wording.sh

 .PHONY: log-analyzer-rules-check
@@ -117,6 +117,9 @@ jobs:
      - name: Check s3s footprint ratchet
        run: ./scripts/check_s3s_footprint.sh

+      - name: Check cryptographic capability wording
+        run: ./scripts/check_fips_wording.sh
+
      - name: Check no planning docs committed
        run: ./scripts/check_no_planning_docs.sh

@@ -152,6 +152,9 @@ jobs:
      - name: Check s3s footprint ratchet
        run: ./scripts/check_s3s_footprint.sh

+      - name: Check cryptographic capability wording
+        run: ./scripts/check_fips_wording.sh
+
      - name: Check no planning docs committed
        run: ./scripts/check_no_planning_docs.sh

@@ -15,28 +15,35 @@
 # Package Workflow - Build DEB/RPM packages
 #
 # This workflow builds DEB and RPM packages from pre-built Linux binaries
-# and uploads them to Cloudflare R2.
+# and uploads them to Cloudflare R2 and the GitHub release.
 #
 # Trigger:
-# - release published: automatically package when a GitHub release is published
-# - workflow_dispatch: manual trigger with optional tag/run_id
+# - workflow_run: automatically package after "Build and Release" completes
+#   for a release tag (the mac/windows/linux binaries are already uploaded
+#   to the GitHub release before packaging starts)
+# - workflow_dispatch: manual fallback (backfill / re-run) with optional tag/run_id
 #
 # Flow:
-# 1. Find the Build workflow run for the release tag
+# 1. Resolve the triggering Build workflow run for the release tag
 # 2. Download Linux binaries (x86_64-gnu, aarch64-gnu) from build artifacts
 # 3. Build DEB packages for amd64 and arm64
 # 4. Build RPM packages for x86_64 and aarch64
-# 5. Upload all packages to Cloudflare R2
+# 5. Upload all packages to Cloudflare R2 and the GitHub release

 name: Package DEB/RPM

 permissions:
-  contents: read
+  # contents: write is required to upload packages to the GitHub release
+  contents: write
  actions: read

 on:
-  release:
-    types: [ published ]
+  # Follows the same pattern as docker.yml: run after the release build
+  # workflow completes, so packaging is triggered only by release tags
+  # (e.g. 1.0.0-rc.2, 1.0.0-rc.3), never by development builds.
+  workflow_run:
+    workflows: [ "Build and Release" ]
+    types: [ completed ]
  workflow_dispatch:
    inputs:
      tag:
@@ -49,13 +56,26 @@ on:
        type: string

 concurrency:
-  group: ${{ github.workflow }}-${{ github.event.release.tag_name || github.event.inputs.tag || github.run_id }}
+  group: ${{ github.workflow }}-${{ github.event.workflow_run.head_branch || github.event.inputs.tag || github.run_id }}
  cancel-in-progress: true

+env:
+  HEAD_BRANCH: ${{ github.event.workflow_run.head_branch }}
+  WORKFLOW_RUN_ID: ${{ github.event.workflow_run.id }}
+
 jobs:
  # Resolve which build run to use and extract version info
  resolve:
    name: Resolve Build
+    # Auto-trigger only from successful tag builds of "Build and Release".
+    # Tag pushes arrive as event == push with head_branch != main (a
+    # non-main push head_branch is the release tag name). Manual dispatch
+    # stays available as a fallback for backfills and re-runs.
+    if: >-
+      github.event_name == 'workflow_dispatch' ||
+      (github.event.workflow_run.conclusion == 'success' &&
+       github.event.workflow_run.event == 'push' &&
+       github.event.workflow_run.head_branch != 'main')
    runs-on: ubuntu-latest
    timeout-minutes: 10
    outputs:
@@ -75,8 +95,8 @@ jobs:
          set -euo pipefail

          # Determine tag
-          if [[ "${{ github.event_name }}" == "release" ]]; then
-            TAG="${{ github.event.release.tag_name }}"
+          if [[ "${{ github.event_name }}" == "workflow_run" ]]; then
+            TAG="${HEAD_BRANCH}"
          elif [[ -n "$INPUT_TAG" ]]; then
            TAG="$INPUT_TAG"
          else
@@ -93,6 +113,11 @@ jobs:
            BUILD_RUN_ID="$INPUT_RUN_ID"
            echo "Using explicit build run ID: $BUILD_RUN_ID"

+          elif [[ "${{ github.event_name }}" == "workflow_run" ]]; then
+            # Use the Build and Release run that triggered this workflow
+            BUILD_RUN_ID="${WORKFLOW_RUN_ID}"
+            echo "Using triggering workflow run: $BUILD_RUN_ID"
+
          elif [[ -n "$TAG" ]]; then
            # Find the build run that produced this tag
            echo "Looking for build run for tag: $TAG"
@@ -456,6 +481,54 @@ jobs:
            echo "✅ Latest packages updated"
          fi

+      - name: Upload packages to GitHub Release
+        if: needs.resolve.outputs.tag != ''
+        env:
+          GH_TOKEN: ${{ github.token }}
+        shell: bash
+        run: |
+          set -euo pipefail
+
+          TAG="${{ needs.resolve.outputs.tag }}"
+          DEB_FILE="${{ steps.deb.outputs.deb_file }}"
+          RPM_FILE="${{ steps.rpm.outputs.rpm_file }}"
+
+          # Upload the packages, then refresh the release checksums so the new
+          # assets are covered, matching the binary release flow.
+          for f in "$DEB_FILE" "$RPM_FILE"; do
+            if [[ -n "$f" && -f "$f" ]]; then
+              echo "📤 Uploading $(basename "$f") to GitHub release ${TAG}..."
+              gh release upload "$TAG" "$f" --clobber
+            fi
+          done
+
+          CHECKSUM_DIR="$(mktemp -d)"
+          gh release download "$TAG" -p 'SHA256SUMS' -p 'SHA512SUMS' \
+            -D "$CHECKSUM_DIR" --clobber 2>/dev/null || true
+
+          for spec in "SHA256SUMS:sha256sum" "SHA512SUMS:sha512sum"; do
+            asset="${spec%%:*}"
+            checksum_cmd="${spec##*:}"
+            checksum_file="${CHECKSUM_DIR}/${asset}"
+
+            touch "$checksum_file"
+
+            for f in "$DEB_FILE" "$RPM_FILE"; do
+              if [[ -n "$f" && -f "$f" ]]; then
+                base="$(basename "$f")"
+                # Remove any stale entry, then append the fresh digest
+                grep -Fv -- "$base" "$checksum_file" > "${checksum_file}.tmp" || true
+                mv "${checksum_file}.tmp" "$checksum_file"
+                (cd "$(dirname "$f")" && "$checksum_cmd" -- "$base") >> "$checksum_file"
+              fi
+            done
+
+            echo "📤 Updating ${asset} for release ${TAG}..."
+            gh release upload "$TAG" "$checksum_file" --clobber
+          done
+
+          echo "✅ GitHub release assets updated"
+
  # Summary
  summary:
    name: Summary
@@ -31,7 +31,7 @@ HTTP request
      → storage/ecfs (erasure coding, encryption, checksums)
        → ecstore (disk pool selection, data distribution)
          → rio (reader pipeline: encrypt → compress → hash → write)
-            → io-core (zero-copy I/O, buffer pool, direct I/O)
+            → io-core (buffer pool, storage profiling, admission control)
              → local disk / remote disk via RPC
 ```

@@ -55,7 +55,7 @@ rustfs/                      # Workspace root (virtual manifest)
 ├── crates/                  # library crates (authoritative list: Cargo.toml [workspace].members)
 │   ├── ecstore/             # Erasure-coded storage engine
 │   ├── rio/                 # Reader I/O pipeline (encrypt, compress, hash)
-│   ├── io-core/             # Zero-copy I/O, scheduling, buffer pool
+│   ├── io-core/             # Buffer pool, storage profiling, admission control
 │   ├── io-metrics/          # I/O metrics collection
 │   ├── common/              # Shared runtime state, globals, data usage types
 │   ├── config/              # Configuration types and parsing
@@ -302,7 +302,7 @@ The binary (`main.rs`) boots in this order:
              │                  │                  │
        ┌─────▼──────┐    ┌──────▼──────┐    ┌──────▼──────┐
        │  ecstore   │    │     rio     │    │   io-core   │
-        │   (core)   │    │  (readers)  │    │ (zero-copy) │
+        │   (core)   │    │  (readers)  │    │ (buffers)   │
        └─────┬──────┘    └─────────────┘    └─────────────┘
              │
     ┌─────┬──┼──┬─────┬──────┐
@@ -314,7 +314,7 @@ The binary (`main.rs`) boots in this order:

 - **"Where does S3 PutObject go?"**
  `server/` routes → `app/object_usecase` validates → `storage/ecfs` encodes →
-  `ecstore` distributes → `rio` encrypts/compresses → `io-core` writes
+  `ecstore` distributes → `rio` encrypts/compresses → `io-core` supplies buffers

 - **"Where are bucket policies enforced?"**
  `app/bucket_usecase` calls into `crates/policy/`
@@ -964,9 +964,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-kms"
-version = "1.114.0"
+version = "1.115.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c0b7d906608ee41e7ddea9983577ba82200435644d567d63dc34e822e088b453"
+checksum = "d5b034f8b7ceadb873d0bc607c30bb4b0be68e09a84c837174e7c2c6878ff882"
 dependencies = [
 "arc-swap",
 "aws-credential-types",
@@ -990,9 +990,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-s3"
-version = "1.141.0"
+version = "1.142.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d9f9420d3a2467eed22ed3635ca653653162c386a0b0f65c78189f9bd3c1379e"
+checksum = "f9e15a5c55e05f4b0b7e483160b3c85cccdf77cff02c95504f3e71d460855cd2"
 dependencies = [
 "arc-swap",
 "aws-credential-types",
@@ -1027,9 +1027,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-sso"
-version = "1.105.0"
+version = "1.106.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ffd0fbe7873cb548a7aa60f9573c268fff94155397fd4f14dc9f1ecaaab8516"
+checksum = "2d0efcee834347b6705eca3eea2defd88242f43774f55d7326604222e3c86260"
 dependencies = [
 "arc-swap",
 "aws-credential-types",
@@ -1053,9 +1053,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-ssooidc"
-version = "1.107.0"
+version = "1.108.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "175763eb222a46377df7aa257a3bca980ab3e96703fefc8f4d0b8da6ad2e254c"
+checksum = "a59312a04cf19c962cfee32b64ecfee758f8786407ff6da5b30fff46ae96f201"
 dependencies = [
 "arc-swap",
 "aws-credential-types",
@@ -1079,9 +1079,9 @@ dependencies = [

 [[package]]
 name = "aws-sdk-sts"
-version = "1.110.0"
+version = "1.111.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dd8b14781dfbff48984017d57167b6ea0b6471c6920ec52b44a2677c7feb3c13"
+checksum = "120e7eb63457a9e547f9986fe3b273f77c43679da4d04f46359fa881c5e19b6e"
 dependencies = [
 "arc-swap",
 "aws-credential-types",
@@ -1598,7 +1598,7 @@ version = "0.10.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3078c7629b62d3f0439517fa394996acacc5cbc91c5a20d8c658e77abd503a71"
 dependencies = [
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 ]

 [[package]]
@@ -1617,7 +1617,7 @@ version = "0.3.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a8894febbff9f758034a5b8e12d87918f56dfc64a8e1fe757d65e29041538d93"
 dependencies = [
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 ]

 [[package]]
@@ -1858,9 +1858,9 @@ dependencies = [

 [[package]]
 name = "cc"
-version = "1.4.2"
+version = "1.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "5d262e149917187838d5b42777c8253bcb64500067342904e7d429499a6f277e"
+checksum = "509591b7bcd67f4ef775afad7662703b4935daaa6ec0e5605cfb1090b32a2b6d"
 dependencies = [
 "find-msvc-tools",
 "jobserver",
@@ -1968,7 +1968,7 @@ version = "0.4.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "773f3b9af64447d2ce9850330c473515014aa235e6a783b02db81ff39e4a3dad"
 dependencies = [
- "crypto-common 0.1.6",
+ "crypto-common 0.1.7",
 "inout 0.1.4",
 ]

@@ -2428,7 +2428,7 @@ version = "0.5.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0dc92fb57ca44df6db8059111ab3af99a63d5d0f8375d9972e319a379c6bab76"
 dependencies = [
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 "rand_core 0.6.4",
 "subtle",
 "zeroize",
@@ -2453,11 +2453,11 @@ dependencies = [

 [[package]]
 name = "crypto-common"
-version = "0.1.6"
+version = "0.1.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1bfb12502f3fc46cca1bb51ac28df9d618d813cdc3d2f25b9fe775a34af26bb3"
+checksum = "78c8292055d1c1df0cce5d180393dc8cce0abec0a7102adb6c7b1eef6016d60a"
 dependencies = [
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 "typenum",
 ]

@@ -3664,7 +3664,7 @@ checksum = "9ed9a281f7bc9b7576e61468ba615a66a5c8cfdff42420a70aa82701a3b1e292"
 dependencies = [
 "block-buffer 0.10.4",
 "const-oid 0.9.6",
- "crypto-common 0.1.6",
+ "crypto-common 0.1.7",
 "subtle",
 ]

@@ -3924,7 +3924,7 @@ dependencies = [
 "crypto-bigint 0.5.5",
 "digest 0.10.7",
 "ff 0.13.1",
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 "group 0.13.0",
 "hkdf 0.12.4",
 "pem-rfc7468 0.7.0",
@@ -4148,9 +4148,9 @@ dependencies = [

 [[package]]
 name = "find-msvc-tools"
-version = "0.1.10"
+version = "0.1.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "26b73573e6edcd2af0cdf47bd6cb58f0b3839491263c314eaad1ccf24430e1de"
+checksum = "d45db016d36b838f563236e9193d0ee6ce38f3f68b6c94e914b4929c96bbb890"

 [[package]]
 name = "findshlibs"
@@ -4369,9 +4369,9 @@ dependencies = [

 [[package]]
 name = "generic-array"
-version = "0.14.9"
+version = "0.14.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4bb6743198531e02858aeaea5398fcc883e71851fcbcb5a2f773e2fb6cb1edf2"
+checksum = "85649ca51fd72272d7821adaf274ad91c288277713d9c18820d8499a7ff69e9a"
 dependencies = [
 "typenum",
 "version_check",
@@ -4380,11 +4380,11 @@ dependencies = [

 [[package]]
 name = "generic-array"
-version = "1.4.4"
+version = "1.4.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab4e5aa225bc56696909483320f0ff9b600f1a971b52e07a17d70f3d9b43254b"
+checksum = "337d46834ee672ab3e48caca2cb0c78cc174fb12b3a68d0d88f99a0519a5e36e"
 dependencies = [
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 "rustversion",
 "typenum",
 ]
@@ -4726,9 +4726,9 @@ dependencies = [

 [[package]]
 name = "h2"
-version = "0.4.15"
+version = "0.4.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6cb093c84e8bd9b188d4c4a8cb6579fc016968d14c99882163cd3ff402a4f155"
+checksum = "a9f37a958b41b3b19ee2707c06439c0e9e547e847223eb791ecb0cb821c65e27"
 dependencies = [
 "atomic-waker",
 "bytes",
@@ -5028,9 +5028,9 @@ dependencies = [

 [[package]]
 name = "hotpath"
-version = "0.23.2"
+version = "0.23.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "62e810bedda5a467ef5c9b5c8a20763fefebc89b63ef36f7ee44a143085204a2"
+checksum = "dce755d457a63bdd0c95e4c91511daad1b58b33209543b7f38027b676f387e5e"
 dependencies = [
 "arc-swap",
 "async-channel",
@@ -5062,9 +5062,9 @@ dependencies = [

 [[package]]
 name = "hotpath-macros"
-version = "0.23.2"
+version = "0.23.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "01bdc59bfc1a9984bee2ff5da63b2f6fccbaa57cd9a4119d709524632bddf341"
+checksum = "a903af89a8429cb07790c3818bc15270b394f80af1bc254e5ccf9c7de2961770"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -5073,15 +5073,15 @@ dependencies = [

 [[package]]
 name = "hotpath-macros-meta"
-version = "0.23.2"
+version = "0.23.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d9216e8a01abe1e1671c376dc8736fb1bf772d7a889538d25f9e1200120ced38"
+checksum = "bcc0ab94ffbb2ee77f4a897df02b5a137a10cf24d69bda936e59aff4dd456e61"

 [[package]]
 name = "hotpath-meta"
-version = "0.23.2"
+version = "0.23.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f22a9d20435fb79511b19dae37b3607224cd98f342a410702d84657cc38fc72f"
+checksum = "053481f6cec8f775a3276c7f6e2f21123111d28261e4edc15ea7421c445964bb"
 dependencies = [
 "hotpath-macros-meta",
 ]
@@ -5280,9 +5280,9 @@ dependencies = [

 [[package]]
 name = "icu_collections"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2984d1cd16c883d7935b9e07e44071dca8d917fd52ecc02c04d5fa0b5a3f191c"
+checksum = "fa68d21081c4a05d5a901a1c62add574c77048b6a1c67be3b50ce0b60d4ca513"
 dependencies = [
 "displaydoc",
 "potential_utf",
@@ -5294,9 +5294,9 @@ dependencies = [

 [[package]]
 name = "icu_locale_core"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "92219b62b3e2b4d88ac5119f8904c10f8f61bf7e95b640d25ba3075e6cac2c29"
+checksum = "d56e28588da92eee5c3201a6eff33fabdd49b62269c8938d4ff050ce4d900deb"
 dependencies = [
 "displaydoc",
 "litemap",
@@ -5307,9 +5307,9 @@ dependencies = [

 [[package]]
 name = "icu_normalizer"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c56e5ee99d6e3d33bd91c5d85458b6005a22140021cc324cea84dd0e72cff3b4"
+checksum = "12f9cf5f235641ed274641dd81c3f28d870e276763d0797aeeab72317b1c646f"
 dependencies = [
 "icu_collections",
 "icu_normalizer_data",
@@ -5321,16 +5321,17 @@ dependencies = [

 [[package]]
 name = "icu_normalizer_data"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "da3be0ae77ea334f4da67c12f149704f19f81d1adf7c51cf482943e84a2bad38"
+checksum = "1563da1ed3e0b3bf3d74c9b85917ac9c56464d2f57242270c09c9e752f8021a0"

 [[package]]
 name = "icu_properties"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bee3b67d0ea5c2cca5003417989af8996f8604e34fb9ddf96208a033901e70de"
+checksum = "7e7ca276ad3145661a65914e6daf131ca5120cd3dcee8f8f3214b8875184a148"
 dependencies = [
+ "displaydoc",
 "icu_collections",
 "icu_locale_core",
 "icu_properties_data",
@@ -5341,15 +5342,15 @@ dependencies = [

 [[package]]
 name = "icu_properties_data"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8e2bbb201e0c04f7b4b3e14382af113e17ba4f63e2c9d2ee626b720cbce54a14"
+checksum = "e590f038c1464a96894fd6d10127e90a8be4509f56ff7ecef851b15cee0b7caa"

 [[package]]
 name = "icu_provider"
-version = "2.2.0"
+version = "2.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "139c4cf31c8b5f33d7e199446eff9c1e02decfc2f0eec2c8d71f65befa45b421"
+checksum = "92a7ed671a6aad807a8651a2e1782a6598fda9ce5185dd8158549e95a91c6428"
 dependencies = [
 "displaydoc",
 "icu_locale_core",
@@ -5417,7 +5418,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "879f10e63c20629ecabbb64a8010319738c66a5cd0c29b02d63d272b03751d01"
 dependencies = [
 "block-padding 0.3.3",
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 ]

 [[package]]
@@ -5968,9 +5969,9 @@ dependencies = [

 [[package]]
 name = "libredox"
-version = "0.1.19"
+version = "0.1.20"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2026a5056764a10b2bf5d56488cba40da507f5493a6a429340e2004d9ed085fa"
+checksum = "28d0a00925a9f930d679b6789b721e3a7f9ed110f41b86d2497caa780c3a070a"
 dependencies = [
 "libc",
 ]
@@ -6033,9 +6034,9 @@ checksum = "32a66949e030da00e8c7d4434b251670a91556f4144941d37452769c25d58a53"

 [[package]]
 name = "litemap"
-version = "0.8.2"
+version = "0.8.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "92daf443525c4cce67b150400bc2316076100ce0b3686209eb8cf3c31612e6f0"
+checksum = "47d9d19d1d6efa0109d2f65ff4c85cddd50bd572e5a00127ab10987290bcefae"

 [[package]]
 name = "local-ip-address"
@@ -6482,9 +6483,9 @@ dependencies = [

 [[package]]
 name = "mqttbytes-core-next"
-version = "0.33.3"
+version = "0.34.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3ff7ae19c74aba9e0ed6e4071cd52aa364e020076fa3cc6ef17e43662f756f3c"
+checksum = "366b6ba2b4209ca4bc5ac731ccddf570d09831981eed07e5fbd63564cf0cf1aa"
 dependencies = [
 "bytes",
 "thiserror 2.0.20",
@@ -6885,7 +6886,7 @@ version = "5.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "51e219e79014df21a225b1860a479e2dcd7cbd9130f4defd4bd0e191ea31d67d"
 dependencies = [
- "base64 0.21.7",
+ "base64 0.22.1",
 "chrono",
 "getrandom 0.2.17",
 "http 1.5.0",
@@ -7344,9 +7345,9 @@ dependencies = [

 [[package]]
 name = "pageant"
-version = "0.2.1"
+version = "0.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4f3a5ae18f65a85c67a77d18d42d3606c07948e3c17c1e5f74852b26589e88a5"
+checksum = "3adadc44070da6f464b0918655a12f5792c156e088d8c4082d13e27d94c3e791"
 dependencies = [
 "base16ct 1.0.0",
 "byteorder",
@@ -7728,9 +7729,9 @@ dependencies = [

 [[package]]
 name = "pkg-config"
-version = "0.3.33"
+version = "0.3.34"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "19f132c84eca552bf34cab8ec81f1c1dcc229b811638f9d283dceabe58c5569e"
+checksum = "f6b464fbc74e149a392436b17d523f769e057cb6877f6a5c4618bc6f11800548"

 [[package]]
 name = "plotters"
@@ -7836,9 +7837,9 @@ dependencies = [

 [[package]]
 name = "potential_utf"
-version = "0.1.5"
+version = "0.1.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0103b1cef7ec0cf76490e969665504990193874ea05c85ff9bab8b911d0a0564"
+checksum = "d83eb9bc6d8e5cf568e7a1101d60ee05e81ed50ea106026f3d18deeb046d7661"
 dependencies = [
 "zerovec",
 ]
@@ -8046,7 +8047,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "be769465445e8c1474e9c5dac2018218498557af32d9ed057325ec9a41ae81bf"
 dependencies = [
 "heck 0.5.0",
- "itertools 0.10.5",
+ "itertools 0.14.0",
 "log",
 "multimap",
 "once_cell",
@@ -8066,7 +8067,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "03da047801ff44bb6a4d407d4860c05fd70bb81714e6b2f3812603d5b145b042"
 dependencies = [
 "heck 0.5.0",
- "itertools 0.10.5",
+ "itertools 0.14.0",
 "log",
 "multimap",
 "petgraph 0.8.3",
@@ -8087,7 +8088,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8a56d757972c98b346a9b766e3f02746cde6dd1cd1d1d563472929fdd74bec4d"
 dependencies = [
 "anyhow",
- "itertools 0.10.5",
+ "itertools 0.14.0",
 "proc-macro2",
 "quote",
 "syn 2.0.119",
@@ -8100,7 +8101,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b570b25f7617e43d59005d0990ccb79e950a423952cea19671b7a876da390adf"
 dependencies = [
 "anyhow",
- "itertools 0.10.5",
+ "itertools 0.14.0",
 "proc-macro2",
 "quote",
 "syn 2.0.119",
@@ -8216,7 +8217,7 @@ dependencies = [
 "reqwest",
 "serde_json",
 "smallvec",
- "spin 0.12.2",
+ "spin 0.12.3",
 "symbolic-demangle",
 "tempfile",
 "thiserror 2.0.20",
@@ -8285,9 +8286,9 @@ dependencies = [

 [[package]]
 name = "quinn-proto"
-version = "0.11.16"
+version = "0.11.17"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2f4bfc015262b9df63c8845072ce59068853ff5872180c2ce2f13038b970e560"
+checksum = "04759210543be93709136e28212294a659ef5001836ff4eab4d663e4529bba83"
 dependencies = [
 "aws-lc-rs",
 "bytes",
@@ -8553,9 +8554,9 @@ dependencies = [

 [[package]]
 name = "redis"
-version = "1.5.0"
+version = "1.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3257df217f7eab0044627a268c9cc6cdb60c0c421c88f83ac41c4e31520b6b84"
+checksum = "e37a4ca5c6ca42aa3e6df2fd32b987a65d32a4c2159a6f3fe0fd1df306a2658f"
 dependencies = [
 "arc-swap",
 "arcstr",
@@ -8567,7 +8568,7 @@ dependencies = [
 "futures-channel",
 "futures-util",
 "itoa",
- "num-bigint 0.4.8",
+ "num-bigint 0.5.1",
 "percent-encoding",
 "pin-project-lite",
 "rustls",
@@ -8868,9 +8869,9 @@ dependencies = [

 [[package]]
 name = "rumqttc-core-next"
-version = "0.33.3"
+version = "0.34.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7d7d9205738dd41a2546e82d27a634d07d8b303dcf7558565ff70caf3ceb0f9c"
+checksum = "249896ab27ed630590971738264baa8f722f18965d2e387c706c40a3c2a572cc"
 dependencies = [
 "async-tungstenite",
 "futures-io",
@@ -8886,18 +8887,18 @@ dependencies = [

 [[package]]
 name = "rumqttc-next"
-version = "0.33.3"
+version = "0.34.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ed1bad2180ff539da671da9a996152a921bc5316eb6d8a9cc3bd441653138b08"
+checksum = "477c9bbfba8f3aecc7aad31c6de2eacb75822efaa18e7aeecb8d3d8e534fbf07"
 dependencies = [
 "rumqttc-v5-next",
 ]

 [[package]]
 name = "rumqttc-v5-next"
-version = "0.33.3"
+version = "0.34.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "229576cbedfa9089f90c17c9454e9429ac1e89cdd223bac5cb39d837593f79bc"
+checksum = "3dfa6ddcc7a7dd5688f9bf78d8f81cb94f367bce56c055d8d94cf81ecb0518bf"
 dependencies = [
 "async-tungstenite",
 "bytes",
@@ -8920,9 +8921,9 @@ dependencies = [

 [[package]]
 name = "russh"
-version = "0.62.6"
+version = "0.62.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b41043523e0edcbd4e31d00903e26f12994f63b21bae9904f7405c1ed92752a5"
+checksum = "9decb68e4e44e1079700e54f17c8f23806ec53d7e0db73ab1c71d9dabc666812"
 dependencies = [
 "aes 0.9.2",
 "aws-lc-rs",
@@ -8945,7 +8946,7 @@ dependencies = [
 "enum_dispatch",
 "flate2",
 "futures",
- "generic-array 1.4.4",
+ "generic-array 1.4.5",
 "getrandom 0.4.3",
 "ghash",
 "hex-literal",
@@ -9280,6 +9281,7 @@ dependencies = [
 "s3s",
 "serde",
 "serde_json",
+ "smallvec",
 "tokio",
 "tonic",
 "tracing",
@@ -9463,6 +9465,7 @@ dependencies = [
 "tokio-stream",
 "tokio-util",
 "tonic",
+ "tonic-prost",
 "tower",
 "tracing",
 "tracing-core",
@@ -9490,7 +9493,7 @@ dependencies = [
 "parking_lot",
 "rayon",
 "smallvec",
- "spin 0.12.2",
+ "spin 0.12.3",
 ]

 [[package]]
@@ -9536,6 +9539,8 @@ version = "1.0.0-rc.2"
 dependencies = [
 "async-trait",
 "base64 0.23.1",
+ "bytes",
+ "crc-fast",
 "futures",
 "hotpath",
 "http 1.5.0",
@@ -9608,7 +9613,6 @@ version = "1.0.0-rc.2"
 dependencies = [
 "bytes",
 "hotpath",
- "memmap2",
 "rustfs-io-metrics",
 "thiserror 2.0.20",
 "tokio",
@@ -9825,14 +9829,19 @@ name = "rustfs-madmin"
 version = "1.0.0-rc.2"
 dependencies = [
 "hotpath",
+ "http 1.5.0",
 "humantime",
 "hyper",
 "jiff",
+ "reqwest",
 "rmp-serde",
+ "rustfs-signer",
+ "s3s",
 "serde",
 "serde_json",
 "sysinfo",
 "time",
+ "tokio",
 ]

 [[package]]
@@ -10246,6 +10255,7 @@ dependencies = [
 "rustfs-ecstore",
 "rustfs-filemeta",
 "rustfs-lock",
+ "rustfs-s3-types",
 "rustfs-storage-api",
 "rustfs-utils",
 "s3s",
@@ -10826,7 +10836,7 @@ checksum = "d3e97a565f76233a6003f9f5c54be1d9c5bdfa3eccfb189469f11ec4901c47dc"
 dependencies = [
 "base16ct 0.2.0",
 "der 0.7.10",
- "generic-array 0.14.9",
+ "generic-array 0.14.7",
 "pkcs8 0.10.2",
 "subtle",
 "zeroize",
@@ -11390,9 +11400,9 @@ checksum = "023a211cb3138dbc438680b32560ad89f699977624c9f8dbb95a47d5b4c07dd3"

 [[package]]
 name = "spin"
-version = "0.12.2"
+version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8abadc99fd9c7bbb7d0ca2b31d72a067d0c0dcd7aad25ab8cac71ba91417694b"
+checksum = "0134f9043ed38b087ac4f7d4af44c79e2c9e5094421fe3164f435ce585953b10"
 dependencies = [
 "lock_api",
 ]
@@ -11804,7 +11814,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd"
 dependencies = [
 "fastrand",
- "getrandom 0.3.4",
+ "getrandom 0.4.3",
 "once_cell",
 "rustix",
 "windows-sys 0.61.2",
@@ -11968,9 +11978,9 @@ dependencies = [

 [[package]]
 name = "tinystr"
-version = "0.8.3"
+version = "0.8.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8323304221c2a851516f22236c5722a72eaa19749016521d6dff0824447d96d"
+checksum = "b1e27c91459209c2986af3dcf603a5a74a4368754ce37414f59acc971167f643"
 dependencies = [
 "displaydoc",
 "zerovec",
@@ -12644,9 +12654,9 @@ checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"

 [[package]]
 name = "uuid"
-version = "1.24.0"
+version = "1.24.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bf3923a6f5c4c6382e0b653c4117f48d631ea17f38ed86e2a828e6f7412f5239"
+checksum = "2cefc03fd367c0c6d4305de1b312cf00248c4114f4a0418ce6a6af769e3b0bd9"
 dependencies = [
 "getrandom 0.4.3",
 "js-sys",
@@ -13161,9 +13171,9 @@ dependencies = [

 [[package]]
 name = "writeable"
-version = "0.6.3"
+version = "0.6.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1ffae5123b2d3fc086436f8834ae3ab053a283cfac8fe0a0b8eaae044768a4c4"
+checksum = "3ad82d2a33cdc9674dc7465672f271e096168fcdbe0f799d9e6db8c5892679dc"

 [[package]]
 name = "x509-cert"
@@ -13343,9 +13353,9 @@ dependencies = [

 [[package]]
 name = "zerotrie"
-version = "0.2.4"
+version = "0.2.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0f9152d31db0792fa83f70fb2f83148effb5c1f5b8c7686c3459e361d9bc20bf"
+checksum = "4ea269c3bd32f0a32c321907a2ae912ba6f4649bb0fc764a15627e99a7095a3f"
 dependencies = [
 "displaydoc",
 "yoke",
@@ -13354,9 +13364,9 @@ dependencies = [

 [[package]]
 name = "zerovec"
-version = "0.11.6"
+version = "0.11.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "90f911cbc359ab6af17377d242225f4d75119aec87ea711a880987b18cd7b239"
+checksum = "94b5c6b5976d66c1d703c4fd17d3f5e43c8cedaacf604961b171adc7130896d8"
 dependencies = [
 "yoke",
 "zerofrom",
@@ -13365,13 +13375,13 @@ dependencies = [

 [[package]]
 name = "zerovec-derive"
-version = "0.11.3"
+version = "0.11.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "625dc425cab0dca6dc3c3319506e6593dcb08a9f387ea3b284dbd52a92c40555"
+checksum = "9f212a141d820099d57ffafb9569be9617a6f27d3dc881fbee8fb56642f917a9"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn 2.0.119",
+ "syn 3.0.3",
 ]

 [[package]]
@@ -228,9 +228,9 @@ atoi = "3.1.0"
 atomic_enum = "0.3.0"
 aws-config = { version = "1.10.1" }
 aws-credential-types = { version = "1.3.0" }
-aws-sdk-kms = { default-features = false, version = "1.114.0" }
-aws-sdk-s3 = { default-features = false, version = "1.141.0" }
-aws-sdk-sts = { default-features = false, version = "1.110.0" }
+aws-sdk-kms = { default-features = false, version = "1.115.0" }
+aws-sdk-s3 = { default-features = false, version = "1.142.0" }
+aws-sdk-sts = { default-features = false, version = "1.111.0" }
 aws-smithy-http-client = { default-features = false, version = "1.3.0" }
 aws-smithy-runtime-api = { version = "1.14.0" }
 aws-smithy-types = { version = "1.6.2" }
@@ -284,8 +284,8 @@ rayon = "1.12.0"
 reed-solomon-erasure = { package = "rustfs-erasure-codec", version = "8.0.2" }
 reed-solomon-simd = "3.1.0"
 regex = { version = "1.13.1" }
-rumqttc = { package = "rumqttc-next", version = "0.33.3" }
-redis = { version = "1.5.0" }
+rumqttc = { package = "rumqttc-next", version = "0.34.0" }
+redis = { version = "1.6.0" }
 rustify = { version = "0.7", default-features = false }
 rustix = { version = "1.1.4" }
 rust-embed = { version = "8.12.0" }
@@ -313,7 +313,7 @@ tracing-subscriber = { version = "0.3.23" }
 transform-stream = "0.3.1"
 url = "2.5.8"
 urlencoding = "2.1.3"
-uuid = { version = "1.24.0" }
+uuid = { version = "1.24.1" }
 vaultrs = { version = "0.8.0" }
 tar = "0.4.46"
 walkdir = "2.5.0"
@@ -341,7 +341,7 @@ libunftp = { version = "0.23.0" }
 unftp-core = "0.1.0"
 suppaftp = { version = "10.0.1" }
 rcgen = { version = "0.14.9", default-features = false, features = ["aws_lc_rs", "crypto", "pem"] }
-russh = { version = "0.62.6" }
+russh = { version = "0.62.7" }
 russh-sftp = "2.4.0"

 # WebDAV
@@ -350,7 +350,7 @@ dav-server = "0.11.0"
 # Performance Analysis and Memory Profiling
 mimalloc = { version = "0.1.52", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11" }
 libmimalloc-sys = { version = "0.1.49", git = "https://github.com/xonatius/mimalloc_rust.git", rev = "6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11", features = ["extended"] }
-hotpath = { version = "0.23.2", default-features = false }
+hotpath = { version = "0.23.3", default-features = false }
 # Snapshot testing for output format regression detection
 insta = { version = "1.48" }

@@ -40,6 +40,7 @@ mak = "mak"
 gae = "gae"
 GAE = "GAE"
 thr = "thr"
+mis = "mis"
 # s3-tests original test names (cannot be changed)
 nonexisted = "nonexisted"
 consts = "consts"
@@ -11,7 +11,6 @@
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#![allow(dead_code)]

 use base64_simd::STANDARD;

@@ -38,7 +38,10 @@ pub const XXHASH_3_HEADER_NAME: &str = "x-amz-checksum-xxhash3";
 pub const XXHASH_64_HEADER_NAME: &str = "x-amz-checksum-xxhash64";
 pub const XXHASH_128_HEADER_NAME: &str = "x-amz-checksum-xxhash128";

-#[allow(dead_code)]
+#[allow(
+    dead_code,
+    reason = "Content-MD5 wire name, resolved by header_name() below and asserted by this crate's tests (backlog#1823)"
+)]
 pub(crate) static MD5_HEADER_NAME: &str = "content-md5";

 pub const CHECKSUM_ALGORITHMS_IN_PRIORITY_ORDER: [&str; 5] =
@@ -476,13 +476,19 @@ impl Checksum for Xxhash64 {
    }
 }

-#[allow(dead_code)]
 #[derive(Debug, Default)]
+#[allow(
+    dead_code,
+    reason = "Content-MD5 is not a ChecksumAlgorithm variant and has no arm in into_impl: S3 carries it as its own header, separate from the x-amz-checksum-* family. This impl exists so the two paths share the Checksum trait, and is asserted by this crate's tests (backlog#1823)"
+)]
 struct Md5 {
    hasher: md5::Md5,
 }

-#[allow(dead_code)]
+#[allow(
+    dead_code,
+    reason = "Content-MD5 is not a ChecksumAlgorithm variant and has no arm in into_impl: S3 carries it as its own header, separate from the x-amz-checksum-* family. This impl exists so the two paths share the Checksum trait, and is asserted by this crate's tests (backlog#1823)"
+)]
 impl Md5 {
    fn update(&mut self, bytes: &[u8]) {
        use md5::Digest;
@@ -42,6 +42,7 @@ chrono = { workspace = true, features = ["serde"] }
 jiff = { workspace = true, features = ["serde"] }
 metrics = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
+smallvec = { workspace = true }
 rmp-serde = { workspace = true }
 s3s = { workspace = true, features = ["minio"] }
 tracing = { workspace = true }
@@ -224,6 +224,13 @@ pub struct HealOpts {
 pub enum HealAdmissionDropReason {
    QueueFull,
    PolicyDropped,
+    /// HS-06: an admin heal start overlaps (same bucket with mutually
+    /// containing prefixes, or the same erasure set) an already running or
+    /// queued task. Only produced when RUSTFS_HEAL_OVERLAP_POLICY=minio_error.
+    AlreadyRunning,
+    /// HS-06: same as [`Self::AlreadyRunning`] but for paths that merely
+    /// contain (or are contained by) the active task's path.
+    OverlappingPaths,
 }

 impl HealAdmissionDropReason {
@@ -231,6 +238,8 @@ impl HealAdmissionDropReason {
        match self {
            Self::QueueFull => "queue_full",
            Self::PolicyDropped => "policy_dropped",
+            Self::AlreadyRunning => "already_running",
+            Self::OverlappingPaths => "overlapping_paths",
        }
    }
 }
@@ -287,6 +296,9 @@ pub enum HealRequestSource {
    Scanner,
    AutoHeal,
    ReadRepair,
+    /// Mission Repair Feed: intents delivered by error paths and replayed
+    /// from the durable MRF journal.
+    Mrf,
 }

 impl HealRequestSource {
@@ -297,6 +309,7 @@ impl HealRequestSource {
            Self::Scanner => "scanner",
            Self::AutoHeal => "auto_heal",
            Self::ReadRepair => "read_repair",
+            Self::Mrf => "mrf",
        }
    }
 }
@@ -313,6 +326,9 @@ pub enum HealChannelCommand {
    Query {
        heal_path: String,
        client_token: String,
+        /// Incremental result cursor (HS-06): only items with a sequence
+        /// greater than this are returned; `None` keeps the full snapshot.
+        since_seq: Option<u64>,
        response_tx: oneshot::Sender<Result<HealChannelResponse, String>>,
    },
    /// Cancel heal task
@@ -518,10 +534,21 @@ async fn receive_heal_channel_response(

 /// Send heal query request
 pub async fn query_heal_status(heal_path: String, client_token: String) -> Result<HealChannelResponse, String> {
+    query_heal_status_since(heal_path, client_token, None).await
+}
+
+/// Incremental heal query (HS-06): pass the client's last seen sequence
+/// number to receive only newer result items.
+pub async fn query_heal_status_since(
+    heal_path: String,
+    client_token: String,
+    since_seq: Option<u64>,
+) -> Result<HealChannelResponse, String> {
    let (response_tx, response_rx) = oneshot::channel();
    send_heal_command(HealChannelCommand::Query {
        heal_path,
        client_token,
+        since_seq,
        response_tx,
    })
    .await?;
@@ -17,8 +17,10 @@ pub mod globals;
 pub mod heal_channel;
 pub mod last_minute;
 pub mod metrics;
+pub mod mrf_channel;
 mod readiness;
 pub mod table_catalog;
+pub mod trace_bus;

 pub use globals::*;
 pub use readiness::{GlobalReadiness, SystemStage};
@@ -0,0 +1,203 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+//! Mission Repair Feed (MRF) intent channel.
+//!
+//! Producers on error paths (read decode failure, scanner metadata
+//! corruption, partial-write recovery) hand a lightweight [`MrfIntent`] to the
+//! heal crate through a global bounded channel. Delivery is strictly
+//! non-blocking: `try_send_mrf_intent` never awaits and drops the intent
+//! (counting it) when the channel is full or uninitialized — losing one heal
+//! hint is always preferred over stalling an IO path. Durable replay of
+//! unconsumed intents is the consumer's job (see `rustfs-heal`
+//! `heal::mrf_queue`), mirroring MinIO's `.heal/mrf/list.bin`.
+
+use std::sync::{
+    Arc, OnceLock,
+    atomic::{AtomicBool, Ordering},
+};
+use tokio::sync::mpsc;
+use uuid::Uuid;
+
+/// Bounded capacity of the global MRF channel. Backpressure is resolved by
+/// dropping (and counting) intents, never by blocking the producer.
+const MRF_CHANNEL_CAPACITY: usize = 8192;
+
+/// Why an intent was produced. Drives the heal priority mapping on the
+/// consumer side (DecodeFailure -> Urgent, MetadataCorruption -> High,
+/// PartialWrite -> Normal).
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum MrfKind {
+    /// Erasure decode failed while serving a read (read path).
+    DecodeFailure,
+    /// Scanner classified object metadata as corrupt.
+    MetadataCorruption,
+    /// A write left the object with fewer committed shards than the set size.
+    PartialWrite,
+}
+
+impl MrfKind {
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            MrfKind::DecodeFailure => "decode-failure",
+            MrfKind::MetadataCorruption => "metadata-corruption",
+            MrfKind::PartialWrite => "partial-write",
+        }
+    }
+}
+
+/// One repair intent. Kept deliberately small so the in-memory queue and the
+/// journal stay bounded; `bucket`/`object` are `Arc<str>` so re-arming an
+/// intent never re-allocates the strings.
+#[derive(Clone, Debug)]
+pub struct MrfIntent {
+    pub bucket: Arc<str>,
+    pub object: Arc<str>,
+    /// Version the intent targets, as raw UUID bytes.
+    pub version_id: Option<[u8; 16]>,
+    pub kind: MrfKind,
+    pub enqueued_at_ms: u64,
+    /// Times this intent has already been offered to the heal manager.
+    /// Dropped by the consumer once it reaches `MRF_MAX_ATTEMPTS`.
+    pub attempts: u8,
+}
+
+/// Consumer-side retry ceiling before an intent is given up on.
+pub const MRF_MAX_ATTEMPTS: u8 = 3;
+
+impl MrfIntent {
+    /// Rough in-memory footprint used by the queue's byte budget.
+    pub fn estimated_bytes(&self) -> usize {
+        // Struct + strings + version bytes; buckets and objects are usually
+        // far below this bound, so rounding up keeps the budget conservative.
+        64 + self.bucket.len() + self.object.len()
+    }
+}
+
+static GLOBAL_MRF_SENDER: OnceLock<mpsc::Sender<MrfIntent>> = OnceLock::new();
+
+/// Delivery kill-switch, set from `RUSTFS_HEAL_MRF_ENABLE`. Producers check
+/// this before touching the channel so the disabled path stays allocation- and
+/// sync-free.
+static MRF_DELIVERY_ENABLED: AtomicBool = AtomicBool::new(true);
+
+/// Override delivery (used at heal-runtime startup from configuration).
+pub fn set_mrf_delivery_enabled(enabled: bool) {
+    MRF_DELIVERY_ENABLED.store(enabled, Ordering::Relaxed);
+}
+
+/// Whether producers currently deliver intents.
+pub fn mrf_delivery_enabled() -> bool {
+    MRF_DELIVERY_ENABLED.load(Ordering::Relaxed)
+}
+
+/// Create the global MRF channel and return the consumer half. Fails if the
+/// channel is already initialized (the heal runtime is a singleton).
+pub fn init_mrf_channel() -> Result<mpsc::Receiver<MrfIntent>, &'static str> {
+    let (sender, receiver) = mpsc::channel(MRF_CHANNEL_CAPACITY);
+    GLOBAL_MRF_SENDER
+        .set(sender)
+        .map_err(|_| "MRF channel sender already initialized")?;
+    Ok(receiver)
+}
+
+/// Best-effort, non-blocking intent delivery from an error path.
+///
+/// Returns `true` when the intent was accepted into the channel. `false`
+/// means the intent was dropped (feature disabled, channel not yet
+/// initialized, or channel full) — callers must not retry or await; the
+/// existing read-repair / scanner heal paths remain the safety net.
+///
+/// This runs on IO error paths, so it stays synchronous and cheap: one
+/// bounded allocation for the two `Arc<str>` handles plus the channel slot.
+pub fn try_send_mrf_intent(kind: MrfKind, bucket: &str, object: &str, version_id: Option<Uuid>) -> bool {
+    if !mrf_delivery_enabled() {
+        return false;
+    }
+    let Some(sender) = GLOBAL_MRF_SENDER.get() else {
+        return false;
+    };
+    let intent = MrfIntent {
+        bucket: Arc::from(bucket),
+        object: Arc::from(object),
+        version_id: version_id.map(|vid| *vid.as_bytes()),
+        kind,
+        enqueued_at_ms: unix_now_ms(),
+        attempts: 0,
+    };
+    sender.try_send(intent).is_ok()
+}
+
+fn unix_now_ms() -> u64 {
+    // Kept trivial: the timestamp is diagnostic metadata only; wall-clock
+    // failure would be a bug rather than something to handle here.
+    std::time::SystemTime::now()
+        .duration_since(std::time::UNIX_EPOCH)
+        .map(|d| d.as_millis() as u64)
+        .unwrap_or(0)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn intents_estimate_is_conservative() {
+        let intent = MrfIntent {
+            bucket: Arc::from("bucket"),
+            object: Arc::from("object"),
+            version_id: Some([0u8; 16]),
+            kind: MrfKind::DecodeFailure,
+            enqueued_at_ms: 0,
+            attempts: 0,
+        };
+        assert!(intent.estimated_bytes() >= intent.bucket.len() + intent.object.len());
+    }
+
+    #[tokio::test]
+    async fn try_send_delivers_and_respects_capacity() {
+        let mut receiver = init_mrf_channel().expect("first initialization should succeed");
+        assert!(init_mrf_channel().is_err(), "double initialization must fail");
+
+        assert!(try_send_mrf_intent(MrfKind::DecodeFailure, "b", "o", Some(Uuid::nil())));
+        let intent = receiver.recv().await.expect("intent should arrive");
+        assert_eq!(intent.kind, MrfKind::DecodeFailure);
+        assert_eq!(intent.bucket.as_ref(), "b");
+
+        // Disable delivery: producers become no-ops.
+        set_mrf_delivery_enabled(false);
+        assert!(!try_send_mrf_intent(MrfKind::PartialWrite, "b", "o", None));
+        set_mrf_delivery_enabled(true);
+
+        // Fill the bounded channel past capacity: excess intents are dropped,
+        // never blocking.
+        let mut accepted = 0;
+        for _ in 0..(MRF_CHANNEL_CAPACITY + 64) {
+            if try_send_mrf_intent(MrfKind::PartialWrite, "b", "o", None) {
+                accepted += 1;
+            }
+        }
+        assert_eq!(accepted, MRF_CHANNEL_CAPACITY);
+    }
+
+    #[test]
+    fn try_send_without_channel_is_false() {
+        // This test may run after the tokio test above in the same process;
+        // the singleton semantics make a clean "uninitialized" case hard, so
+        // assert the flag-off behavior only.
+        set_mrf_delivery_enabled(false);
+        assert!(!try_send_mrf_intent(MrfKind::MetadataCorruption, "b", "o", None));
+        set_mrf_delivery_enabled(true);
+    }
+}
@@ -0,0 +1,333 @@
+// Copyright 2024 RustFS Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use smallvec::SmallVec;
+use std::{
+    sync::{
+        Arc, OnceLock,
+        atomic::{AtomicUsize, Ordering},
+    },
+    time::{Duration, SystemTime},
+};
+use tokio::sync::broadcast;
+
+const DEFAULT_TRACE_BUS_CAPACITY: usize = 1024;
+const TRACE_ATTR_INLINE_CAPACITY: usize = 8;
+
+static GLOBAL_TRACE_BUS: OnceLock<TraceBus> = OnceLock::new();
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum TraceKind {
+    Heal,
+    Scanner,
+}
+
+impl TraceKind {
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::Heal => "heal",
+            Self::Scanner => "scanner",
+        }
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum TraceFunc {
+    HealTask,
+    HealBucket,
+    HealObject,
+    HealCheckAbandonedParts,
+    HealErasureSetPage,
+    ScannerFolder,
+    ScannerIlmAction,
+    ScannerHealCandidate,
+    Dropped,
+}
+
+impl TraceFunc {
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::HealTask => "heal.Task",
+            Self::HealBucket => "heal.Bucket",
+            Self::HealObject => "heal.Object",
+            Self::HealCheckAbandonedParts => "heal.CheckAbandonedParts",
+            Self::HealErasureSetPage => "heal.ErasureSetPage",
+            Self::ScannerFolder => "scanner.Folder",
+            Self::ScannerIlmAction => "scanner.IlmAction",
+            Self::ScannerHealCandidate => "scanner.HealCandidate",
+            Self::Dropped => "trace.Dropped",
+        }
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum TraceVal {
+    Bool(bool),
+    U64(u64),
+    I64(i64),
+    Str(Arc<str>),
+}
+
+impl From<bool> for TraceVal {
+    fn from(value: bool) -> Self {
+        Self::Bool(value)
+    }
+}
+
+impl From<u64> for TraceVal {
+    fn from(value: u64) -> Self {
+        Self::U64(value)
+    }
+}
+
+impl From<i64> for TraceVal {
+    fn from(value: i64) -> Self {
+        Self::I64(value)
+    }
+}
+
+impl From<&str> for TraceVal {
+    fn from(value: &str) -> Self {
+        Self::Str(Arc::from(value))
+    }
+}
+
+impl From<String> for TraceVal {
+    fn from(value: String) -> Self {
+        Self::Str(Arc::from(value))
+    }
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct TraceAttr {
+    pub key: &'static str,
+    pub value: TraceVal,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct TraceEvent {
+    pub kind: TraceKind,
+    pub func: TraceFunc,
+    pub time: SystemTime,
+    pub bucket: Option<Arc<str>>,
+    pub object: Option<Arc<str>>,
+    pub duration: Duration,
+    pub bytes: u64,
+    pub attrs: SmallVec<[TraceAttr; TRACE_ATTR_INLINE_CAPACITY]>,
+}
+
+impl TraceEvent {
+    pub fn new(kind: TraceKind, func: TraceFunc) -> Self {
+        Self {
+            kind,
+            func,
+            time: SystemTime::now(),
+            bucket: None,
+            object: None,
+            duration: Duration::ZERO,
+            bytes: 0,
+            attrs: SmallVec::new(),
+        }
+    }
+
+    pub fn with_bucket(mut self, bucket: impl Into<Arc<str>>) -> Self {
+        self.bucket = Some(bucket.into());
+        self
+    }
+
+    pub fn with_object(mut self, object: impl Into<Arc<str>>) -> Self {
+        self.object = Some(object.into());
+        self
+    }
+
+    pub fn with_duration(mut self, duration: Duration) -> Self {
+        self.duration = duration;
+        self
+    }
+
+    pub fn with_bytes(mut self, bytes: u64) -> Self {
+        self.bytes = bytes;
+        self
+    }
+
+    pub fn with_attr(mut self, key: &'static str, value: impl Into<TraceVal>) -> Self {
+        self.attrs.push(TraceAttr {
+            key,
+            value: value.into(),
+        });
+        self
+    }
+}
+
+#[derive(Debug)]
+pub struct TraceBus {
+    sender: broadcast::Sender<Arc<TraceEvent>>,
+    subscriber_count: Arc<AtomicUsize>,
+}
+
+impl TraceBus {
+    pub fn new(capacity: usize) -> Self {
+        let capacity = capacity.max(1);
+        let (sender, _receiver) = broadcast::channel(capacity);
+        Self {
+            sender,
+            subscriber_count: Arc::new(AtomicUsize::new(0)),
+        }
+    }
+
+    pub fn subscriber_count(&self) -> usize {
+        self.subscriber_count.load(Ordering::Acquire)
+    }
+
+    pub fn subscribe(&self) -> TraceSubscription {
+        let receiver = self.sender.subscribe();
+        self.subscriber_count.fetch_add(1, Ordering::AcqRel);
+        TraceSubscription {
+            receiver,
+            subscriber_count: Arc::clone(&self.subscriber_count),
+        }
+    }
+
+    pub fn emit(&self, build: impl FnOnce() -> TraceEvent) -> bool {
+        if self.subscriber_count() == 0 {
+            return false;
+        }
+
+        self.sender.send(Arc::new(build())).is_ok()
+    }
+}
+
+impl Default for TraceBus {
+    fn default() -> Self {
+        Self::new(DEFAULT_TRACE_BUS_CAPACITY)
+    }
+}
+
+#[derive(Debug)]
+pub struct TraceSubscription {
+    receiver: broadcast::Receiver<Arc<TraceEvent>>,
+    subscriber_count: Arc<AtomicUsize>,
+}
+
+impl TraceSubscription {
+    pub async fn recv(&mut self) -> Result<Arc<TraceEvent>, broadcast::error::RecvError> {
+        self.receiver.recv().await
+    }
+
+    pub fn try_recv(&mut self) -> Result<Arc<TraceEvent>, broadcast::error::TryRecvError> {
+        self.receiver.try_recv()
+    }
+}
+
+impl Drop for TraceSubscription {
+    fn drop(&mut self) {
+        self.subscriber_count.fetch_sub(1, Ordering::AcqRel);
+    }
+}
+
+pub fn global_trace_bus() -> &'static TraceBus {
+    GLOBAL_TRACE_BUS.get_or_init(TraceBus::default)
+}
+
+pub fn subscribe_trace_events() -> TraceSubscription {
+    global_trace_bus().subscribe()
+}
+
+pub fn trace_emit(build: impl FnOnce() -> TraceEvent) -> bool {
+    global_trace_bus().emit(build)
+}
+
+pub fn trace_subscriber_count() -> usize {
+    global_trace_bus().subscriber_count()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::sync::atomic::AtomicUsize;
+
+    #[test]
+    fn trace_emit_skips_builder_without_subscribers() {
+        let bus = TraceBus::new(4);
+        let built = AtomicUsize::new(0);
+
+        let sent = bus.emit(|| {
+            built.fetch_add(1, Ordering::Relaxed);
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealTask)
+        });
+
+        assert!(!sent);
+        assert_eq!(built.load(Ordering::Relaxed), 0);
+    }
+
+    #[tokio::test]
+    async fn trace_subscriber_receives_event() {
+        let bus = TraceBus::new(4);
+        let mut subscription = bus.subscribe();
+
+        assert!(bus.emit(|| {
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealObject)
+                .with_bucket("bucket")
+                .with_object("object")
+                .with_duration(Duration::from_millis(7))
+                .with_bytes(11)
+                .with_attr("dry", true)
+        }));
+
+        let event = subscription
+            .recv()
+            .await
+            .expect("subscriber should receive emitted trace event");
+
+        assert_eq!(event.kind, TraceKind::Heal);
+        assert_eq!(event.func, TraceFunc::HealObject);
+        assert_eq!(event.bucket.as_deref(), Some("bucket"));
+        assert_eq!(event.object.as_deref(), Some("object"));
+        assert_eq!(event.duration, Duration::from_millis(7));
+        assert_eq!(event.bytes, 11);
+        assert_eq!(
+            event.attrs.as_slice(),
+            &[TraceAttr {
+                key: "dry",
+                value: TraceVal::Bool(true)
+            }]
+        );
+    }
+
+    #[test]
+    fn trace_subscription_drop_decrements_count() {
+        let bus = TraceBus::new(4);
+        let subscription = bus.subscribe();
+
+        assert_eq!(bus.subscriber_count(), 1);
+        drop(subscription);
+        assert_eq!(bus.subscriber_count(), 0);
+    }
+
+    #[tokio::test]
+    async fn lagged_subscriber_drops_events_without_blocking_publishers() {
+        let bus = TraceBus::new(2);
+        let mut subscription = bus.subscribe();
+
+        for index in 0_u64..4 {
+            assert!(bus.emit(|| { TraceEvent::new(TraceKind::Scanner, TraceFunc::ScannerFolder).with_attr("index", index) }));
+        }
+
+        let err = subscription
+            .recv()
+            .await
+            .expect_err("receiver should observe lag instead of blocking publishers");
+        assert!(matches!(err, broadcast::error::RecvError::Lagged(_)));
+    }
+}
@@ -14,9 +14,8 @@

 //! Shared backpressure policy type.
 //!
-//! The runtime backpressure implementation (byte-watermark pipes and
-//! monitors) lives in `rustfs/src/storage/backpressure.rs`; this module only
-//! carries the watermark policy type that implementation shares.
+//! This module only carries the watermark policy; the admission primitive it
+//! projects into lives in `rustfs-io-core`.

 use rustfs_io_core::BackpressureConfig as CoreBackpressureConfig;

@@ -177,3 +177,40 @@ pub const DEFAULT_HEAL_MAINLINE_WRITE_UTILIZATION_HIGH_PERCENT: usize = 80;

 /// Default foreground pressure recheck delay for heal scheduler, in milliseconds.
 pub const DEFAULT_HEAL_MAINLINE_MAX_SLEEP_MS: u64 = 250;
+
+/// Environment variable that toggles the MRF (mission repair feed) intent
+/// pipeline: error paths deliver repair intents to the heal runtime, and
+/// unconsumed intents are replayed from the durable journal after a restart.
+pub const ENV_HEAL_MRF_ENABLE: &str = "RUSTFS_HEAL_MRF_ENABLE";
+
+/// Environment variable for the MRF in-memory queue capacity (intent count).
+pub const ENV_HEAL_MRF_QUEUE_SIZE: &str = "RUSTFS_HEAL_MRF_QUEUE_SIZE";
+
+/// Environment variable for the MRF journal byte budget. The journal is
+/// compacted once its on-disk size crosses this bound.
+pub const ENV_HEAL_MRF_JOURNAL_MAX_BYTES: &str = "RUSTFS_HEAL_MRF_JOURNAL_MAX_BYTES";
+
+/// Environment variable for the MRF journal replay batch size (intents per
+/// replay push round).
+pub const ENV_HEAL_MRF_REPLAY_BATCH: &str = "RUSTFS_HEAL_MRF_REPLAY_BATCH";
+
+/// Default behavior keeps the MRF intent pipeline enabled.
+pub const DEFAULT_HEAL_MRF_ENABLE: bool = true;
+
+/// Default MRF queue capacity (matches MinIO's 100k MRF list ceiling).
+pub const DEFAULT_HEAL_MRF_QUEUE_SIZE: usize = 100_000;
+
+/// Default MRF journal byte budget (8 MiB), mirroring the channel payload cap.
+pub const DEFAULT_HEAL_MRF_JOURNAL_MAX_BYTES: usize = 8 * 1024 * 1024;
+
+/// Default MRF replay batch size.
+pub const DEFAULT_HEAL_MRF_REPLAY_BATCH: usize = 256;
+
+/// Environment variable selecting how admin heal starts behave when the
+/// requested path overlaps an already running or queued heal: `merge`
+/// (default, keep today's dedup/merge semantics) or `minio_error` (return a
+/// typed already-running / overlapping-paths rejection like madmin).
+pub const ENV_HEAL_OVERLAP_POLICY: &str = "RUSTFS_HEAL_OVERLAP_POLICY";
+
+/// Default overlap policy: merge duplicate/overlapping requests.
+pub const DEFAULT_HEAL_OVERLAP_POLICY: &str = "merge";
@@ -234,6 +234,31 @@ pub const ENV_OBJECT_DISK_WRITE_ABSOLUTE_CAP: &str = "RUSTFS_OBJECT_DISK_WRITE_A
 /// Default absolute per-object erasure write cap in seconds (`0` = disabled).
 pub const DEFAULT_OBJECT_DISK_WRITE_ABSOLUTE_CAP: u64 = 0;

+/// Enable foreground PutObject request admission.
+///
+/// This is an experimental, default-off foreground write backpressure gate for
+/// strict commit tail investigations. When disabled, PUTs follow the legacy
+/// path and only the existing request counters are updated.
+pub const ENV_PUT_FOREGROUND_ADMISSION_ENABLE: &str = "RUSTFS_PUT_FOREGROUND_ADMISSION_ENABLE";
+pub const DEFAULT_PUT_FOREGROUND_ADMISSION_ENABLE: bool = false;
+
+/// Maximum foreground PutObject requests admitted concurrently per process.
+///
+/// The limit is used only when [`ENV_PUT_FOREGROUND_ADMISSION_ENABLE`] is true.
+/// A value of `0` disables the gate even when the enable flag is present, so a
+/// partially configured rollout cannot reject every PUT.
+pub const ENV_PUT_FOREGROUND_ADMISSION_LIMIT: &str = "RUSTFS_PUT_FOREGROUND_ADMISSION_LIMIT";
+pub const DEFAULT_PUT_FOREGROUND_ADMISSION_LIMIT: usize = 0;
+
+/// Time in milliseconds a foreground PutObject waits for an admission permit.
+///
+/// Once this timeout expires the request fails before body ingest/storage
+/// mutation with S3 `SlowDown`/503. `0` means fail fast when the limit is full.
+pub const ENV_PUT_FOREGROUND_ADMISSION_WAIT_TIMEOUT_MS: &str = "RUSTFS_PUT_FOREGROUND_ADMISSION_WAIT_TIMEOUT_MS";
+pub const DEFAULT_PUT_FOREGROUND_ADMISSION_WAIT_TIMEOUT_MS: u64 = 0;
+
+const _: () = assert!(!DEFAULT_PUT_FOREGROUND_ADMISSION_ENABLE);
+
 /// Environment variable for minimum GetObject timeout in seconds.
 ///
 /// When dynamic timeout calculation is enabled, this is the minimum timeout
@@ -870,6 +870,157 @@ pub struct DataUsageCacheInfo {
    pub snapshot_complete: bool,
 }

+/// Prefix-level usage over a raw entry map — the shared core behind
+/// [`DataUsageCache::prefix_usage`], usable by any cache-shaped reader (the
+/// scanner's writer-side cache has the same map type).
+///
+/// Cache keys are cleaned literal paths (`bucket/pre/fix`), so sub-prefix
+/// names come straight off the child keys — no reverse mapping exists or is
+/// needed. A compacted prefix carries its aggregate but no children, which
+/// the `compacted` flag reports so callers can say why the breakdown is
+/// empty. `truncated` is set when the breakdown exceeded `max_entries` and
+/// was cut (largest first).
+pub fn prefix_usage_in_cache(
+    cache: &HashMap<String, DataUsageEntry>,
+    bucket: &str,
+    prefix: &str,
+    max_entries: usize,
+) -> Option<PrefixUsageQuery> {
+    let prefix = prefix.trim_matches('/');
+    let root = if prefix.is_empty() {
+        bucket.to_string()
+    } else {
+        format!("{bucket}/{prefix}")
+    };
+    let entry = cache.get(&hash_path(&root).key())?.clone();
+
+    let usage = PrefixUsageSummary::from_entry(&flatten_entry(cache, &entry, 0)?);
+
+    let child_prefix = format!("{root}/");
+    let mut sub_prefixes: Vec<PrefixUsageEntry> = entry
+        .children
+        .iter()
+        .filter_map(|child_key| {
+            let child = cache.get(child_key)?;
+            let child_flat = flatten_entry(cache, child, 1)?;
+            // Child keys are literal `bucket/pre/name` paths; a trailing
+            // slash marks a directory object and is display-only here.
+            let name = child_key
+                .strip_prefix(child_prefix.as_str())
+                .unwrap_or(child_key.as_str())
+                .trim_end_matches('/')
+                .to_string();
+            Some(PrefixUsageEntry {
+                prefix: name,
+                usage: PrefixUsageSummary::from_entry(&child_flat),
+            })
+        })
+        .collect();
+    sub_prefixes.sort_by(|left, right| {
+        right
+            .usage
+            .size
+            .cmp(&left.usage.size)
+            .then_with(|| left.prefix.cmp(&right.prefix))
+    });
+    let truncated = sub_prefixes.len() > max_entries;
+    sub_prefixes.truncate(max_entries);
+
+    Some(PrefixUsageQuery {
+        usage,
+        compacted: entry.compacted,
+        truncated,
+        sub_prefixes,
+    })
+}
+
+/// Maximum subtree depth [`flatten_entry`] will walk before declaring the
+/// cache corrupt — the same bound the scanner's checked flatten uses.
+const PREFIX_USAGE_MAX_DEPTH: usize = 1024;
+
+/// Flatten one entry's subtree into an aggregate: the free-function twin of
+/// [`DataUsageCache::flatten`], carrying the scanner checked-flatten
+/// hardening so a corrupt cache (cycles, over-deep trees, overflowing
+/// counters) yields `None` instead of unbounded recursion or wrapped totals.
+fn flatten_entry(cache: &HashMap<String, DataUsageEntry>, root: &DataUsageEntry, depth: usize) -> Option<DataUsageEntry> {
+    if depth > PREFIX_USAGE_MAX_DEPTH {
+        return None;
+    }
+    let mut flattened = DataUsageEntry::default();
+    if !flattened.checked_merge(root) {
+        return None;
+    }
+    flattened.compacted = root.compacted;
+    // The root itself is not pre-seeded: it is merged above, and a corrupt
+    // child edge pointing back at the root's own key is still terminated by
+    // the visited set on first encounter.
+    let mut visited: HashSet<&str> = HashSet::new();
+    let mut pending: Vec<(&String, usize)> = root.children.iter().map(|child| (child, depth + 1)).collect();
+    while let Some((key, child_depth)) = pending.pop() {
+        if child_depth > PREFIX_USAGE_MAX_DEPTH || !visited.insert(key.as_str()) {
+            return None;
+        }
+        let entry = cache.get(key)?;
+        if !flattened.checked_merge(entry) {
+            return None;
+        }
+        pending.extend(entry.children.iter().map(|child| (child, child_depth + 1)));
+    }
+    flattened.children.clear();
+    Some(flattened)
+}
+
+/// Flattened counters of one prefix subtree, as returned by
+/// [`DataUsageCache::prefix_usage`].
+#[derive(Clone, Copy, Debug, Default, PartialEq, Eq, serde::Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct PrefixUsageSummary {
+    pub size: u64,
+    pub objects: u64,
+    pub versions: u64,
+    pub delete_markers: u64,
+}
+
+impl PrefixUsageSummary {
+    fn from_entry(entry: &DataUsageEntry) -> Self {
+        Self {
+            size: entry.size as u64,
+            objects: entry.objects as u64,
+            versions: entry.versions as u64,
+            delete_markers: entry.delete_markers as u64,
+        }
+    }
+
+    /// Add another set's counters into this one (entries are partitioned by
+    /// set, so per-set results sum).
+    pub fn merge(&mut self, other: &Self) {
+        self.size = self.size.saturating_add(other.size);
+        self.objects = self.objects.saturating_add(other.objects);
+        self.versions = self.versions.saturating_add(other.versions);
+        self.delete_markers = self.delete_markers.saturating_add(other.delete_markers);
+    }
+}
+
+/// One first-level sub-prefix row of a [`PrefixUsageQuery`].
+#[derive(Clone, Debug, PartialEq, Eq, serde::Serialize)]
+pub struct PrefixUsageEntry {
+    pub prefix: String,
+    pub usage: PrefixUsageSummary,
+}
+
+/// Result of [`DataUsageCache::prefix_usage`].
+#[derive(Clone, Debug, Default, PartialEq, Eq, serde::Serialize)]
+#[serde(rename_all = "camelCase")]
+pub struct PrefixUsageQuery {
+    pub usage: PrefixUsageSummary,
+    /// The prefix entry was compacted by the scanner: its aggregate is valid
+    /// but no sub-prefix breakdown exists on disk.
+    pub compacted: bool,
+    /// The breakdown had more entries than `max_entries`; the largest remain.
+    pub truncated: bool,
+    pub sub_prefixes: Vec<PrefixUsageEntry>,
+}
+
 /// Read-only projection of a scanner-written `.usage-cache.bin` file.
 ///
 /// The scanner-side `DataUsageCache` (`crates/scanner/src/data_usage_define.rs`)
@@ -997,6 +1148,21 @@ impl DataUsageCache {
        }
    }

+    /// Prefix-level usage for one bucket subtree, plus the one-level
+    /// breakdown below it (rustfs/backlog#1872, MinIO
+    /// `loadPrefixUsageFromBackend` parity and beyond: arbitrary prefixes and
+    /// full counters instead of first-level sizes only).
+    ///
+    /// Cache keys are cleaned literal paths (`bucket/pre/fix`), so sub-prefix
+    /// names come straight off the child keys — no reverse mapping exists or
+    /// is needed. A compacted prefix carries its aggregate but no children,
+    /// which the `compacted` flag reports so callers can say why the
+    /// breakdown is empty. `truncated` is set when the breakdown exceeded
+    /// `max_entries` and was cut (largest first).
+    pub fn prefix_usage(&self, bucket: &str, prefix: &str, max_entries: usize) -> Option<PrefixUsageQuery> {
+        prefix_usage_in_cache(&self.cache, bucket, prefix, max_entries)
+    }
+
    pub fn force_compact(&mut self, limit: usize) {
        if self.cache.len() < limit {
            return;
@@ -1898,6 +2064,126 @@ mod tests {
        );
    }

+    /// Build a cache shaped like `bucket/{a,b/{c,d}},bucket/loose` with
+    /// distinct counters so aggregation is observable.
+    fn prefix_usage_fixture_cache() -> DataUsageCache {
+        let mut cache = DataUsageCache::default();
+        let mut insert = |path: &str, parent: &str, size: usize, objects: usize, versions: usize, delete_markers: usize| {
+            cache.replace(
+                path,
+                parent,
+                DataUsageEntry {
+                    size,
+                    objects,
+                    versions,
+                    delete_markers,
+                    ..Default::default()
+                },
+            );
+        };
+        insert("bucket", "", 0, 0, 0, 0);
+        insert("bucket/a", "bucket", 100, 1, 1, 0);
+        insert("bucket/b", "bucket", 0, 0, 0, 0);
+        insert("bucket/b/c", "bucket/b", 200, 2, 2, 1);
+        insert("bucket/b/d", "bucket/b", 40, 1, 3, 0);
+        insert("bucket/loose", "bucket", 10, 1, 1, 1);
+        cache
+    }
+
+    #[test]
+    fn prefix_usage_aggregates_bucket_root_and_one_level_below() {
+        let cache = prefix_usage_fixture_cache();
+
+        let root = cache
+            .prefix_usage("bucket", "", 100)
+            .expect("root query must find the bucket entry");
+        assert_eq!(root.usage.size, 350, "root aggregate flattens the whole subtree");
+        assert_eq!(root.usage.objects, 5);
+        assert_eq!(root.usage.versions, 7);
+        assert_eq!(root.usage.delete_markers, 2);
+        assert!(!root.compacted);
+        assert!(!root.truncated);
+        // Breakdown is one level: b (240) before a (100) before loose (10),
+        // each flattened to its own subtree total.
+        let names: Vec<(&str, u64)> = root
+            .sub_prefixes
+            .iter()
+            .map(|entry| (entry.prefix.as_str(), entry.usage.size))
+            .collect();
+        assert_eq!(names, vec![("b", 240), ("a", 100), ("loose", 10)]);
+    }
+
+    #[test]
+    fn prefix_usage_drills_into_arbitrary_prefixes() {
+        let cache = prefix_usage_fixture_cache();
+
+        let b = cache.prefix_usage("bucket", "b", 100).expect("nested prefix must resolve");
+        assert_eq!(b.usage.size, 240);
+        assert_eq!(b.usage.versions, 5);
+        let names: Vec<&str> = b.sub_prefixes.iter().map(|entry| entry.prefix.as_str()).collect();
+        assert_eq!(names, vec!["c", "d"]);
+
+        // Prefix slashes are normalized away.
+        let slashed = cache.prefix_usage("bucket", "/b/", 100).expect("slash-insensitive lookup");
+        assert_eq!(slashed.usage.size, 240);
+
+        assert!(cache.prefix_usage("bucket", "absent", 100).is_none(), "unknown prefix must be a miss");
+        assert!(cache.prefix_usage("other", "", 100).is_none(), "unknown bucket must be a miss");
+    }
+
+    #[test]
+    fn prefix_usage_reports_and_respects_truncation() {
+        let cache = prefix_usage_fixture_cache();
+        let capped = cache.prefix_usage("bucket", "", 2).expect("root query");
+        assert!(capped.truncated, "three children capped to two must flag truncation");
+        let names: Vec<&str> = capped.sub_prefixes.iter().map(|entry| entry.prefix.as_str()).collect();
+        assert_eq!(names, vec!["b", "a"], "largest prefixes survive the cut");
+    }
+
+    #[test]
+    fn prefix_usage_marks_compacted_entries() {
+        let mut cache = DataUsageCache::default();
+        cache.replace(
+            "bucket",
+            "",
+            DataUsageEntry {
+                size: 999,
+                objects: 9,
+                compacted: true,
+                ..Default::default()
+            },
+        );
+
+        let compacted = cache.prefix_usage("bucket", "", 100).expect("compacted root resolves");
+        assert!(compacted.compacted, "compaction must be visible to callers");
+        assert_eq!(compacted.usage.size, 999);
+        assert!(compacted.sub_prefixes.is_empty(), "a compacted entry carries no children");
+    }
+
+    #[test]
+    fn prefix_usage_rejects_cyclic_and_dangling_caches() {
+        // A self-referencing child (corrupt cache) must yield a miss for the
+        // whole query, not unbounded recursion.
+        let mut cache = prefix_usage_fixture_cache();
+        if let Some(entry) = cache.cache.get_mut("bucket/b") {
+            entry.children.insert("bucket/b".to_string());
+        }
+        assert!(cache.prefix_usage("bucket", "b", 100).is_none(), "a cyclic subtree must be rejected");
+        // The unaffected sibling still answers.
+        assert!(cache.prefix_usage("bucket", "a", 100).is_some());
+
+        // A child key with no entry (dangling link) is rejected rather than
+        // silently dropped: half a tree would under-report usage.
+        let mut dangling = prefix_usage_fixture_cache();
+        if let Some(entry) = dangling.cache.get_mut("bucket/b") {
+            entry.children.insert("bucket/b/ghost".to_string());
+        }
+        assert!(
+            dangling.prefix_usage("bucket", "b", 100).is_none(),
+            "a dangling child link must be rejected"
+        );
+    }
+
    #[test]
    fn hash_path_uses_portable_slash_semantics() {
        for (input, expected) in [
@@ -32,6 +32,7 @@ use rustfs_signer::sign_v4;
 use s3s::Body;
 use std::ffi::OsStr;
 use std::fs as stdfs;
+use std::io::ErrorKind;
 use std::path::{Path, PathBuf};
 use std::process::{Child, Command, Stdio};
 use std::sync::Once;
@@ -51,6 +52,11 @@ pub(crate) const FAST_DATA_USAGE_SCANNER_ENV: &[(&str, &str)] =
    &[("RUSTFS_SCANNER_CYCLE", "1"), ("RUSTFS_SCANNER_START_DELAY_SECS", "0")];
 pub const TEST_BUCKET: &str = "e2e-test-bucket";
 const RUSTFS_FULL_FEATURE: &str = "full";
+const TEST_PORT_MIN: u16 = 20_000;
+const TEST_PORT_RANGE: u16 = 40_000;
+const TEST_PORT_COUNTER_PATH: &str = "/tmp/rustfs_e2e_next_port";
+const TEST_PORT_LOCK_DIR: &str = "/tmp/rustfs_e2e_port_allocator.lock";
+const TEST_PORT_LOCK_STALE_AFTER: Duration = Duration::from_secs(30);

 fn capture_log_path(log_dir: &Path, temp_dir: &str) -> Option<PathBuf> {
    let temp_name = Path::new(temp_dir).file_name()?.to_string_lossy();
@@ -67,6 +73,64 @@ fn configured_capture_log_path(temp_dir: &str) -> Option<String> {
    capture_log_path(Path::new(&log_dir), temp_dir).map(|path| path.to_string_lossy().into_owned())
 }

+struct PortAllocatorGuard;
+
+impl PortAllocatorGuard {
+    async fn acquire() -> Result<Self, Box<dyn std::error::Error + Send + Sync>> {
+        loop {
+            match stdfs::create_dir(TEST_PORT_LOCK_DIR) {
+                Ok(()) => return Ok(Self),
+                Err(err) if err.kind() == ErrorKind::AlreadyExists => {
+                    remove_stale_port_allocator_lock();
+                    sleep(Duration::from_millis(10)).await;
+                }
+                Err(err) => return Err(err.into()),
+            }
+        }
+    }
+}
+
+impl Drop for PortAllocatorGuard {
+    fn drop(&mut self) {
+        let _ = stdfs::remove_dir(TEST_PORT_LOCK_DIR);
+    }
+}
+
+fn advance_test_port(port: u16) -> u16 {
+    let offset = (port - TEST_PORT_MIN + 1) % TEST_PORT_RANGE;
+    TEST_PORT_MIN + offset
+}
+
+fn seeded_test_port() -> u16 {
+    let offset = (Uuid::new_v4().as_u128() % u128::from(TEST_PORT_RANGE)) as u16;
+    TEST_PORT_MIN + offset
+}
+
+fn read_next_test_port() -> u16 {
+    stdfs::read_to_string(TEST_PORT_COUNTER_PATH)
+        .ok()
+        .and_then(|value| value.trim().parse::<u16>().ok())
+        .filter(|port| (TEST_PORT_MIN..TEST_PORT_MIN + TEST_PORT_RANGE).contains(port))
+        .unwrap_or_else(seeded_test_port)
+}
+
+fn remove_stale_port_allocator_lock() {
+    let Ok(metadata) = stdfs::metadata(TEST_PORT_LOCK_DIR) else {
+        return;
+    };
+    let Ok(modified) = metadata.modified() else {
+        return;
+    };
+    if modified.elapsed().is_ok_and(|elapsed| elapsed > TEST_PORT_LOCK_STALE_AFTER) {
+        let _ = stdfs::remove_dir(TEST_PORT_LOCK_DIR);
+    }
+}
+
+fn write_next_test_port(port: u16) -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
+    stdfs::write(TEST_PORT_COUNTER_PATH, port.to_string())?;
+    Ok(())
+}
+
 pub(crate) fn capture_command_logs(
    command: &mut Command,
    log_path: Option<&str>,
@@ -508,10 +572,21 @@ impl RustFSTestEnvironment {
    /// Find an available port for the test
    pub async fn find_available_port() -> Result<u16, Box<dyn std::error::Error + Send + Sync>> {
        use std::net::TcpListener;
-        let listener = TcpListener::bind("127.0.0.1:0")?;
-        let port = listener.local_addr()?.port();
-        drop(listener);
-        Ok(port)
+        let _guard = PortAllocatorGuard::acquire().await?;
+        let mut next_port = read_next_test_port();
+
+        for _ in 0..TEST_PORT_RANGE {
+            let port = next_port;
+            next_port = advance_test_port(next_port);
+            write_next_test_port(next_port)?;
+
+            if let Ok(listener) = TcpListener::bind(("127.0.0.1", port)) {
+                drop(listener);
+                return Ok(port);
+            }
+        }
+
+        Err("no available E2E test port found".into())
    }

    /// Kill any existing RustFS processes
@@ -30,7 +30,6 @@ use md5::{Digest as Md5Digest, Md5};
 use rustfs_signer::constants::UNSIGNED_PAYLOAD;
 use rustfs_signer::sign_v4;
 use s3s::Body;
-use serial_test::serial;
 use std::collections::HashMap;
 use std::error::Error;
 use std::io::Cursor;
@@ -356,7 +355,6 @@ async fn run_post_object_policy_case(
 /// smuggles one extra field the policy never declared, and the upload must be
 /// rejected with 403 AccessDenied naming the offending field.
 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_fields_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -484,7 +482,6 @@ async fn test_anonymous_post_object_rejects_fields_missing_from_policy_condition
 /// sends a different one, and the upload must be rejected with 400
 /// InvalidPolicyDocument naming the field.
 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_exact_condition_policy_mismatches()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -689,7 +686,6 @@ async fn test_anonymous_post_object_rejects_exact_condition_policy_mismatches()
 /// one of them with a different value, and the upload must be rejected with
 /// 400 InvalidPolicyDocument naming the mismatched field.
 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_policy_mismatches() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -757,7 +753,6 @@ async fn test_anonymous_post_object_rejects_object_lock_policy_mismatches() -> R
 /// exact values, the form sends a different parameter value, and the upload
 /// must be rejected with 400 InvalidPolicyDocument naming the parameter.
 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sse_kms_policy_mismatches() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -839,7 +834,6 @@ async fn test_anonymous_post_object_rejects_sse_kms_policy_mismatches() -> Resul
 /// NotImplemented (SSE-KMS POST uploads are not implemented), not with a
 /// policy error.
 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sse_kms_params_outside_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -894,7 +888,6 @@ async fn test_anonymous_post_object_rejects_sse_kms_params_outside_policy_condit
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_multipart_control_apis_require_auth() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -968,7 +961,6 @@ async fn test_anonymous_multipart_control_apis_require_auth() -> Result<(), Box<
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_requires_auth() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1002,7 +994,6 @@ async fn test_anonymous_post_object_requires_auth() -> Result<(), Box<dyn std::e
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_honors_success_action_status() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1066,7 +1057,6 @@ async fn test_anonymous_post_object_honors_success_action_status() -> Result<(),
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_honors_success_action_redirect() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1139,7 +1129,6 @@ async fn test_anonymous_post_object_honors_success_action_redirect() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_defaults_to_no_content() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1185,7 +1174,6 @@ async fn test_anonymous_post_object_defaults_to_no_content() -> Result<(), Box<d
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sse_kms() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1232,7 +1220,6 @@ async fn test_anonymous_post_object_rejects_sse_kms() -> Result<(), Box<dyn std:
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_sse_s3() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1290,7 +1277,6 @@ async fn test_anonymous_post_object_accepts_sse_s3() -> Result<(), Box<dyn std::
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_uses_bucket_default_sse_s3() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1363,7 +1349,6 @@ async fn test_anonymous_post_object_uses_bucket_default_sse_s3() -> Result<(), B
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_uses_bucket_default_sse_kms() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1437,7 +1422,6 @@ async fn test_anonymous_post_object_uses_bucket_default_sse_kms() -> Result<(),
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sse_s3_policy_mismatch() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -1488,7 +1472,6 @@ async fn test_anonymous_post_object_rejects_sse_s3_policy_mismatch() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_sse_s3_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1552,7 +1535,6 @@ async fn test_anonymous_post_object_accepts_sse_s3_missing_from_policy_condition
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_storage_class_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1606,7 +1588,6 @@ async fn test_anonymous_post_object_accepts_storage_class_exact_policy_match()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_storage_class_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1657,7 +1638,6 @@ async fn test_anonymous_post_object_rejects_storage_class_missing_from_policy_co
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_invalid_storage_class_value() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -1709,7 +1689,6 @@ async fn test_anonymous_post_object_rejects_invalid_storage_class_value() -> Res
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_checksum_algorithm_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1765,7 +1744,6 @@ async fn test_anonymous_post_object_rejects_checksum_algorithm_missing_from_poli
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_checksum_algorithm_policy_mismatch()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1822,7 +1800,6 @@ async fn test_anonymous_post_object_rejects_checksum_algorithm_policy_mismatch()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_checksum_auxiliary_fields_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1886,7 +1863,6 @@ async fn test_anonymous_post_object_rejects_checksum_auxiliary_fields_missing_fr
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_allows_sse_c_fields_outside_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -1963,7 +1939,6 @@ async fn test_anonymous_post_object_allows_sse_c_fields_outside_policy_condition
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sse_c_exact_policy_mismatch() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -2022,7 +1997,6 @@ async fn test_anonymous_post_object_rejects_sse_c_exact_policy_mismatch() -> Res
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_duplicate_key_form_values() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -2072,7 +2046,6 @@ async fn test_anonymous_post_object_rejects_duplicate_key_form_values() -> Resul
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_invalid_success_action_status() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -2120,7 +2093,6 @@ async fn test_anonymous_post_object_rejects_invalid_success_action_status() -> R
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_invalid_success_action_redirect()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2168,7 +2140,6 @@ async fn test_anonymous_post_object_rejects_invalid_success_action_redirect()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_form_fields_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2223,7 +2194,6 @@ async fn test_anonymous_post_object_rejects_form_fields_missing_from_policy_cond
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_form_fields_covered_by_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2280,7 +2250,6 @@ async fn test_anonymous_post_object_accepts_form_fields_covered_by_policy_condit
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_starts_with_policy_mismatch() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -2335,7 +2304,6 @@ async fn test_anonymous_post_object_rejects_starts_with_policy_mismatch() -> Res
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_content_length_range_violation()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2388,7 +2356,6 @@ async fn test_anonymous_post_object_rejects_content_length_range_violation()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_success_action_status_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2445,7 +2412,6 @@ async fn test_anonymous_post_object_accepts_success_action_status_exact_policy_m
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_success_action_redirect_policy_mismatch()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2502,7 +2468,6 @@ async fn test_anonymous_post_object_rejects_success_action_redirect_policy_misma
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_success_action_redirect_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2568,7 +2533,6 @@ async fn test_anonymous_post_object_accepts_success_action_redirect_exact_policy
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_success_action_redirect_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2621,7 +2585,6 @@ async fn test_anonymous_post_object_rejects_success_action_redirect_missing_from
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_metadata_field_covered_by_starts_with()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2676,7 +2639,6 @@ async fn test_anonymous_post_object_accepts_metadata_field_covered_by_starts_wit
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_content_type_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2734,7 +2696,6 @@ async fn test_anonymous_post_object_accepts_content_type_field_exact_policy_matc
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_content_type_field_covered_by_starts_with()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2792,7 +2753,6 @@ async fn test_anonymous_post_object_accepts_content_type_field_covered_by_starts
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_content_disposition_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2850,7 +2810,6 @@ async fn test_anonymous_post_object_accepts_content_disposition_field_exact_poli
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_cache_control_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2908,7 +2867,6 @@ async fn test_anonymous_post_object_accepts_cache_control_field_exact_policy_mat
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_content_language_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -2966,7 +2924,6 @@ async fn test_anonymous_post_object_accepts_content_language_field_exact_policy_
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_content_encoding_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3024,7 +2981,6 @@ async fn test_anonymous_post_object_accepts_content_encoding_field_exact_policy_
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_website_redirect_location_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3082,7 +3038,6 @@ async fn test_anonymous_post_object_accepts_website_redirect_location_exact_poli
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_expires_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3140,7 +3095,6 @@ async fn test_anonymous_post_object_accepts_expires_field_exact_policy_match()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_retention_without_permission()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3196,7 +3150,6 @@ async fn test_anonymous_post_object_rejects_object_lock_retention_without_permis
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_retention_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3256,7 +3209,6 @@ async fn test_anonymous_post_object_rejects_object_lock_retention_missing_from_p
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_legal_hold_without_permission()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3309,7 +3261,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_without_permi
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_legal_hold_policy_mismatch()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3368,7 +3319,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_policy_mismat
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_object_lock_legal_hold_missing_from_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3426,7 +3376,6 @@ async fn test_anonymous_post_object_rejects_object_lock_legal_hold_missing_from_
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_tagging_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3492,7 +3441,6 @@ async fn test_anonymous_post_object_accepts_tagging_field_exact_policy_match()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_accepts_metadata_field_exact_policy_match()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3551,7 +3499,6 @@ async fn test_anonymous_post_object_accepts_metadata_field_exact_policy_match()
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_allows_x_ignore_fields_outside_policy_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3604,7 +3551,6 @@ async fn test_anonymous_post_object_allows_x_ignore_fields_outside_policy_condit
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_sigv4_date_policy_mismatch() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -3657,7 +3603,6 @@ async fn test_anonymous_post_object_rejects_sigv4_date_policy_mismatch() -> Resu
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_mismatched_bucket_form_field() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -3712,7 +3657,6 @@ async fn test_anonymous_post_object_rejects_mismatched_bucket_form_field() -> Re
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_multiple_bucket_values() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -3764,7 +3708,6 @@ async fn test_anonymous_post_object_rejects_multiple_bucket_values() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_post_object_rejects_extra_content_disposition_field()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3820,7 +3763,6 @@ async fn test_anonymous_post_object_rejects_extra_content_disposition_field()
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_tar_entries_with_prefix_headers()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3891,7 +3833,6 @@ async fn test_signed_put_object_extract_expands_tar_entries_with_prefix_headers(
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_objects()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -3956,7 +3897,6 @@ async fn test_signed_put_object_extract_preserves_request_metadata_on_extracted_
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_sse_s3_and_redirect() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4004,7 +3944,6 @@ async fn test_signed_put_object_extract_preserves_sse_s3_and_redirect() -> Resul
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_storage_class() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4047,7 +3986,6 @@ async fn test_signed_put_object_extract_preserves_storage_class() -> Result<(),
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_rejects_invalid_storage_class() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4083,7 +4021,6 @@ async fn test_signed_put_object_extract_rejects_invalid_storage_class() -> Resul
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_rejects_write_offset_bytes_header() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4137,7 +4074,6 @@ async fn test_signed_put_object_rejects_write_offset_bytes_header() -> Result<()
 }

 #[tokio::test]
-#[serial]
 async fn test_raw_signed_put_object_write_offset_bytes_returns_minio_compatible_error_body()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -4176,7 +4112,6 @@ async fn test_raw_signed_put_object_write_offset_bytes_returns_minio_compatible_
 }

 #[tokio::test]
-#[serial]
 async fn test_anonymous_put_object_write_offset_bytes_returns_minio_compatible_error_body()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -4235,7 +4170,6 @@ async fn test_anonymous_put_object_write_offset_bytes_returns_minio_compatible_e
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_uses_bucket_default_sse_s3() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4300,7 +4234,6 @@ async fn test_signed_put_object_extract_uses_bucket_default_sse_s3() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_rejects_bucket_default_sse_kms() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4356,7 +4289,6 @@ async fn test_signed_put_object_extract_rejects_bucket_default_sse_kms() -> Resu
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_sse_c() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4421,7 +4353,6 @@ async fn test_signed_put_object_extract_preserves_sse_c() -> Result<(), Box<dyn
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_object_lock_legal_hold() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -4476,7 +4407,6 @@ async fn test_signed_put_object_extract_preserves_object_lock_legal_hold() -> Re
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_object_lock_retention() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -4536,7 +4466,6 @@ async fn test_signed_put_object_extract_preserves_object_lock_retention() -> Res
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_pax_retention_overrides_request_retention()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -4600,7 +4529,6 @@ async fn test_signed_put_object_extract_pax_retention_overrides_request_retentio
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_returns_archive_etag() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4634,7 +4562,6 @@ async fn test_signed_put_object_extract_returns_archive_etag() -> Result<(), Box
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_entry_mtime() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -4670,7 +4597,6 @@ async fn test_signed_put_object_extract_preserves_entry_mtime() -> Result<(), Bo
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_pax_metadata_and_version_id()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -4724,7 +4650,6 @@ async fn test_signed_put_object_extract_preserves_pax_metadata_and_version_id()
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_authorizes_each_pax_privilege_and_retention_conditions()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -5034,7 +4959,6 @@ async fn test_signed_put_object_extract_authorizes_each_pax_privilege_and_retent
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_accepts_compat_header() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5076,7 +5000,6 @@ async fn test_signed_put_object_extract_accepts_compat_header() -> Result<(), Bo
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_preserves_directory_markers_by_default()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -5137,7 +5060,6 @@ async fn test_signed_put_object_extract_preserves_directory_markers_by_default()
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_tar_gz_archive() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5189,7 +5111,6 @@ async fn test_signed_put_object_extract_expands_tar_gz_archive() -> Result<(), B
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_tgz_archive() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5241,7 +5162,6 @@ async fn test_signed_put_object_extract_expands_tgz_archive() -> Result<(), Box<
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_tbz2_archive() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5293,7 +5213,6 @@ async fn test_signed_put_object_extract_expands_tbz2_archive() -> Result<(), Box
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_txz_archive() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5345,7 +5264,6 @@ async fn test_signed_put_object_extract_expands_txz_archive() -> Result<(), Box<
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_enabled()
 -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();
@@ -5419,7 +5337,6 @@ async fn test_signed_put_object_extract_skips_invalid_entry_when_ignore_errors_e
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_normalizes_prefix_header_value() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5462,7 +5379,6 @@ async fn test_signed_put_object_extract_normalizes_prefix_header_value() -> Resu
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_expands_tzst_archive() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -5514,7 +5430,6 @@ async fn test_signed_put_object_extract_expands_tzst_archive() -> Result<(), Box
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_rejects_missing_archive_extension() -> Result<(), Box<dyn std::error::Error + Send + Sync>>
 {
    init_logging();
@@ -5548,7 +5463,6 @@ async fn test_signed_put_object_extract_rejects_missing_archive_extension() -> R
 }

 #[tokio::test]
-#[serial]
 async fn test_signed_put_object_extract_rejects_invalid_tar_gz_payload() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
    init_logging();

@@ -33,7 +33,6 @@ use aws_sdk_s3::types::{
    ObjectLockMode, ObjectLockRetentionMode,
 };
 use chrono::{DateTime, Duration, Utc};
-use serial_test::serial;
 use tracing::info;

 /// Initialize test logging
@@ -107,7 +106,6 @@ fn parse_s3_datetime(value: &aws_sdk_s3::primitives::DateTime) -> DateTime<Utc>
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_delete_object_blocked_by_compliance_retention() {
    init_logging();
    info!("🧪 Test: DeleteObject blocked by COMPLIANCE retention");
@@ -145,7 +143,6 @@ async fn test_delete_object_blocked_by_compliance_retention() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_blocked_by_governance_without_bypass() {
    init_logging();
    info!("🧪 Test: DeleteObject blocked by GOVERNANCE retention without bypass");
@@ -175,7 +172,6 @@ async fn test_delete_object_blocked_by_governance_without_bypass() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_allowed_by_governance_with_bypass() {
    init_logging();
    info!("🧪 Test: DeleteObject allowed by GOVERNANCE retention with bypass");
@@ -215,7 +211,6 @@ async fn test_delete_object_allowed_by_governance_with_bypass() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_creates_delete_marker_for_retained_current_version() {
    init_logging();
    info!("🧪 Test: DeleteObject creates delete marker for retained current version");
@@ -266,7 +261,6 @@ async fn test_delete_object_creates_delete_marker_for_retained_current_version()
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_blocked_by_legal_hold() {
    init_logging();
    info!("🧪 Test: DeleteObject blocked by Legal Hold");
@@ -299,7 +293,6 @@ async fn test_delete_object_blocked_by_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_allowed_with_legal_hold_off() {
    init_logging();
    info!("🧪 Test: DeleteObject allowed with Legal Hold OFF");
@@ -335,7 +328,6 @@ async fn test_delete_object_allowed_with_legal_hold_off() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_after_legal_hold_removed() {
    init_logging();
    info!("🧪 Test: DeleteObject succeeds after Legal Hold is removed");
@@ -369,7 +361,6 @@ async fn test_delete_object_after_legal_hold_removed() {
 }

 #[tokio::test]
-#[serial]
 async fn test_get_object_legal_hold_returns_updated_status() {
    init_logging();
    info!("🧪 Test: GetObjectLegalHold returns updated status");
@@ -425,7 +416,6 @@ async fn test_get_object_legal_hold_returns_updated_status() {
 }

 #[tokio::test]
-#[serial]
 async fn test_get_object_retention_returns_configured_values() {
    init_logging();
    info!("🧪 Test: GetObjectRetention returns configured values");
@@ -476,7 +466,6 @@ async fn test_get_object_retention_returns_configured_values() {
 // creating a new current version. The lock protects the existing version
 // from deletion; it never blocks new versions.
 #[tokio::test]
-#[serial]
 async fn test_put_object_overwrite_creates_new_version_under_legal_hold() {
    init_logging();
    info!("🧪 Test: PutObject overwrite of a legal-hold version creates a new version");
@@ -561,7 +550,6 @@ async fn test_put_object_overwrite_creates_new_version_under_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_copy_object_applies_requested_legal_hold() {
    init_logging();
    info!("🧪 Test: CopyObject applies requested Legal Hold");
@@ -613,7 +601,6 @@ async fn test_copy_object_applies_requested_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_copy_object_does_not_inherit_source_legal_hold() {
    init_logging();
    info!("🧪 Test: CopyObject does not inherit source Legal Hold");
@@ -707,7 +694,6 @@ async fn test_copy_object_does_not_inherit_source_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_copy_object_overwrite_creates_new_version_under_legal_hold() {
    init_logging();
    info!("🧪 Test: CopyObject overwrite of a legal-hold destination creates a new version");
@@ -787,7 +773,6 @@ async fn test_copy_object_overwrite_creates_new_version_under_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_create_multipart_upload_applies_requested_legal_hold() {
    init_logging();
    info!("🧪 Test: CreateMultipartUpload applies requested Legal Hold");
@@ -853,7 +838,6 @@ async fn test_create_multipart_upload_applies_requested_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_create_multipart_upload_creates_new_version_under_compliance_retention() {
    init_logging();
    info!("🧪 Test: CreateMultipartUpload over a COMPLIANCE-retained key creates a new version");
@@ -933,7 +917,6 @@ async fn test_create_multipart_upload_creates_new_version_under_compliance_reten
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_completed_multipart_object_blocked_by_legal_hold() {
    init_logging();
    info!("🧪 Test: Delete completed multipart object blocked by Legal Hold");
@@ -993,7 +976,6 @@ async fn test_delete_completed_multipart_object_blocked_by_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_completed_multipart_object_blocked_by_retention() {
    init_logging();
    info!("🧪 Test: Delete completed multipart object blocked by retention");
@@ -1055,7 +1037,6 @@ async fn test_delete_completed_multipart_object_blocked_by_retention() {
 }

 #[tokio::test]
-#[serial]
 async fn test_complete_multipart_upload_creates_new_version_under_legal_hold() {
    init_logging();
    info!("🧪 Test: CompleteMultipartUpload creates a new version when the current version is under Legal Hold");
@@ -1135,7 +1116,6 @@ async fn test_complete_multipart_upload_creates_new_version_under_legal_hold() {
 }

 #[tokio::test]
-#[serial]
 async fn test_complete_multipart_upload_creates_new_version_under_compliance_retention() {
    init_logging();
    info!("🧪 Test: CompleteMultipartUpload creates a new version when the current version is under COMPLIANCE retention");
@@ -1209,7 +1189,6 @@ async fn test_complete_multipart_upload_creates_new_version_under_compliance_ret
 }

 #[tokio::test]
-#[serial]
 async fn test_write_paths_require_put_object_legal_hold_permission() {
    init_logging();
    info!("🧪 Test: write paths require PutObjectLegalHold permission");
@@ -1273,7 +1252,6 @@ async fn test_write_paths_require_put_object_legal_hold_permission() {
 }

 #[tokio::test]
-#[serial]
 async fn test_write_paths_require_put_object_retention_permission() {
    init_logging();
    info!("🧪 Test: write paths require PutObjectRetention permission");
@@ -1345,7 +1323,6 @@ async fn test_write_paths_require_put_object_retention_permission() {
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_delete_objects_mixed_locked_unlocked() {
    init_logging();
    info!("🧪 Test: DeleteObjects with mixed locked and unlocked objects");
@@ -1427,7 +1404,6 @@ async fn test_delete_objects_mixed_locked_unlocked() {
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_put_retention_compliance_cannot_shorten() {
    init_logging();
    info!("🧪 Test: PutObjectRetention cannot shorten COMPLIANCE retention");
@@ -1468,7 +1444,6 @@ async fn test_put_retention_compliance_cannot_shorten() {
 }

 #[tokio::test]
-#[serial]
 async fn test_put_retention_compliance_can_extend() {
    init_logging();
    info!("🧪 Test: PutObjectRetention can extend COMPLIANCE retention");
@@ -1509,7 +1484,6 @@ async fn test_put_retention_compliance_can_extend() {
 }

 #[tokio::test]
-#[serial]
 async fn test_put_retention_governance_extend_without_bypass() {
    init_logging();
    info!("🧪 Test: PutObjectRetention on GOVERNANCE can extend without bypass");
@@ -1553,7 +1527,6 @@ async fn test_put_retention_governance_extend_without_bypass() {
 }

 #[tokio::test]
-#[serial]
 async fn test_put_retention_governance_shorten_requires_bypass() {
    init_logging();
    info!("🧪 Test: PutObjectRetention on GOVERNANCE requires bypass to shorten");
@@ -1615,7 +1588,6 @@ async fn test_put_retention_governance_shorten_requires_bypass() {
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_default_retention_applied_to_new_objects() {
    init_logging();
    info!("🧪 Test: Default retention is applied to new objects");
@@ -1685,7 +1657,6 @@ async fn test_default_retention_applied_to_new_objects() {
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_object_creates_delete_marker_for_default_retained_current_version() {
    init_logging();
    info!("🧪 Test: DeleteObject creates delete marker for default-retained current version");
@@ -1770,7 +1741,6 @@ async fn test_delete_object_creates_delete_marker_for_default_retained_current_v
 }

 #[tokio::test]
-#[serial]
 async fn test_put_copy_and_multipart_reject_incomplete_retention_headers() {
    init_logging();
    info!("🧪 Test: write paths reject incomplete Object Lock retention headers");
@@ -1869,7 +1839,6 @@ async fn test_put_copy_and_multipart_reject_incomplete_retention_headers() {
 }

 #[tokio::test]
-#[serial]
 async fn test_copy_object_retention_uses_destination_policy() {
    init_logging();
    info!("🧪 Test: CopyObject retention follows destination policy");
@@ -2051,7 +2020,6 @@ async fn test_copy_object_retention_uses_destination_policy() {
 }

 #[tokio::test]
-#[serial]
 async fn test_multipart_default_retention_fixed_at_create() {
    init_logging();
    info!("🧪 Test: multipart default retention is fixed at CreateMultipartUpload");
@@ -2122,7 +2090,6 @@ async fn test_multipart_default_retention_fixed_at_create() {
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_unretained_object_lock_object_delete_and_bucket_cleanup() {
    init_logging();
    info!("🧪 Test: Unretained Object Lock object delete and bucket cleanup (Issue #5339)");
@@ -2243,7 +2210,6 @@ async fn test_unretained_object_lock_object_delete_and_bucket_cleanup() {
 }

 #[tokio::test]
-#[serial]
 async fn test_versioning_auto_enabled_with_object_lock() {
    init_logging();
    info!("🧪 Test: Versioning is auto-enabled when Object Lock is configured");
@@ -2302,7 +2268,6 @@ async fn test_versioning_auto_enabled_with_object_lock() {
 // ============================================================================

 #[tokio::test]
-#[serial]
 async fn test_error_message_distinguishes_legal_hold_from_retention() {
    init_logging();
    info!("🧪 Test: Error messages distinguish Legal Hold from Retention");
@@ -60,7 +60,6 @@ use rustfs_signer::constants::UNSIGNED_PAYLOAD;
 use rustfs_signer::sign_v4;
 use s3s::Body;
 use s3s::header::X_AMZ_REPLICATION_STATUS;
-use serial_test::serial;
 use sha2::{Digest, Sha256};
 use std::collections::BTreeMap;
 use std::convert::Infallible;
@@ -2506,7 +2505,6 @@ async fn build_replication_pair(
 /// metadata was inherited wholesale from the source, so the scanner heal pass
 /// skipped it too — no PENDING/FAILED marker meant nothing to re-drive).
 #[tokio::test]
-#[serial]
 async fn test_copy_object_replicates_to_target() -> TestResult {
    init_logging();

@@ -2555,7 +2553,6 @@ async fn test_copy_object_replicates_to_target() -> TestResult {
 /// independent object; every member must replicate to the remote target like a
 /// regular PUT (MinIO PutObjectExtract parity).
 #[tokio::test]
-#[serial]
 async fn test_snowball_extract_replicates_members_to_target() -> TestResult {
    init_logging();

@@ -2601,7 +2598,6 @@ async fn test_snowball_extract_replicates_members_to_target() -> TestResult {
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_succeeds_with_remote_target() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2638,7 +2634,6 @@ async fn test_replication_check_succeeds_with_remote_target() -> Result<(), Box<
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_rejects_target_without_object_lock() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2692,7 +2687,6 @@ async fn test_replication_check_rejects_target_without_object_lock() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_unversioned_source_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2731,7 +2725,6 @@ async fn test_set_remote_target_rejects_unversioned_source_bucket() -> Result<()
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_rejects_unversioned_source_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2755,7 +2748,6 @@ async fn test_replication_check_rejects_unversioned_source_bucket() -> Result<()
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_rejects_missing_replication_config() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2779,7 +2771,6 @@ async fn test_replication_check_rejects_missing_replication_config() -> Result<(
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_rejects_invalid_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2798,7 +2789,6 @@ async fn test_replication_check_rejects_invalid_bucket() -> Result<(), Box<dyn E
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_same_bucket_on_same_deployment() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2842,7 +2832,6 @@ async fn test_set_remote_target_rejects_same_bucket_on_same_deployment() -> Resu
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_unversioned_target_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2876,7 +2865,6 @@ async fn test_set_remote_target_rejects_unversioned_target_bucket() -> Result<()
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_update_requires_arn() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -2928,7 +2916,6 @@ async fn test_set_remote_target_update_requires_arn() -> Result<(), Box<dyn Erro
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_update_rejects_missing_target() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3017,7 +3004,6 @@ async fn fetch_single_target(
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_partial_update_preserves_credentials() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3103,7 +3089,6 @@ async fn test_set_remote_target_partial_update_preserves_credentials() -> Result
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_invalid_target_url() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3145,7 +3130,6 @@ async fn test_set_remote_target_rejects_invalid_target_url() -> Result<(), Box<d
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_self_signed_https_target_without_skip_tls_verify()
 -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();
@@ -3230,7 +3214,6 @@ async fn test_set_remote_target_rejects_self_signed_https_target_without_skip_tl
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_allows_self_signed_https_target_with_skip_tls_verify() -> Result<(), Box<dyn Error + Send + Sync>>
 {
    init_logging();
@@ -3342,7 +3325,6 @@ async fn test_set_remote_target_allows_self_signed_https_target_with_skip_tls_ve
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_pem() -> Result<(), Box<dyn Error + Send + Sync>>
 {
    init_logging();
@@ -3427,7 +3409,6 @@ async fn test_set_remote_target_rejects_private_ca_https_target_without_ca_cert_
 }

 #[tokio::test]
-#[serial]
 async fn test_set_remote_target_allows_private_ca_https_target_with_ca_cert_pem() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3518,7 +3499,6 @@ async fn test_set_remote_target_allows_private_ca_https_target_with_ca_cert_pem(
 }

 #[tokio::test]
-#[serial]
 async fn test_list_remote_targets_rejects_empty_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3538,7 +3518,6 @@ async fn test_list_remote_targets_rejects_empty_bucket() -> Result<(), Box<dyn E
 }

 #[tokio::test]
-#[serial]
 async fn test_list_remote_targets_rejects_invalid_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3557,7 +3536,6 @@ async fn test_list_remote_targets_rejects_invalid_bucket() -> Result<(), Box<dyn
 }

 #[tokio::test]
-#[serial]
 async fn test_remove_remote_target_rejects_missing_target() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3598,7 +3576,6 @@ async fn test_remove_remote_target_rejects_missing_target() -> Result<(), Box<dy
 }

 #[tokio::test]
-#[serial]
 async fn test_remove_remote_target_rejects_missing_arn() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3623,7 +3600,6 @@ async fn test_remove_remote_target_rejects_missing_arn() -> Result<(), Box<dyn E
 }

 #[tokio::test]
-#[serial]
 async fn test_remove_remote_target_rejects_invalid_bucket() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3647,7 +3623,6 @@ async fn test_remove_remote_target_rejects_invalid_bucket() -> Result<(), Box<dy
 }

 #[tokio::test]
-#[serial]
 async fn test_remove_remote_target_rejects_target_used_by_replication() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3687,7 +3662,6 @@ async fn test_remove_remote_target_rejects_target_used_by_replication() -> Resul
 }

 #[tokio::test]
-#[serial]
 async fn test_delete_bucket_replication_removes_remote_target() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3737,7 +3711,6 @@ async fn test_delete_bucket_replication_removes_remote_target() -> Result<(), Bo
 }

 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_replicates_put_object_issue_2539() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -3779,7 +3752,6 @@ async fn test_bucket_replication_replicates_put_object_issue_2539() -> Result<()
 }

 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_converges_delete_marker_and_version_purge() -> TestResult {
    init_logging();

@@ -3878,7 +3850,6 @@ async fn test_bucket_replication_converges_delete_marker_and_version_purge() ->
 }

 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_disabled_delete_marker_does_not_propagate() -> TestResult {
    init_logging();

@@ -3965,7 +3936,6 @@ async fn test_bucket_replication_disabled_delete_marker_does_not_propagate() ->
 /// interoperability profile for a runner that provisions MinIO credentials
 /// and a reachable endpoint.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestResult {
    init_logging();

@@ -4293,7 +4263,6 @@ async fn test_bucket_replication_acceptance_matrix_local_dual_targets() -> TestR
 }

 #[tokio::test]
-#[serial]
 async fn test_single_bucket_multipart_replication_fans_out_to_multiple_targets() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -4462,7 +4431,6 @@ async fn test_repl17_failure_observation_helpers() -> TestResult {
 /// the replica is decryptable only with the original customer key. The
 /// backlog#1291 property still holds: never a silent plaintext replica.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_c_contract() -> TestResult {
    init_logging();

@@ -4540,7 +4508,6 @@ async fn test_bucket_replication_sse_c_contract() -> TestResult {
 /// part — part boundaries and the encrypted-multipart marker survive so the
 /// replica decrypts each part with its part-derived nonce.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_c_multipart_passthrough() -> TestResult {
    init_logging();

@@ -4657,7 +4624,6 @@ async fn test_bucket_replication_sse_c_multipart_passthrough() -> TestResult {
 /// (independent KMS, so success proves target-owned envelopes), preserved
 /// source ETag, and a version that stays stable across scanner cycles.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_s3_contract() -> TestResult {
    init_logging();
    assert_managed_sse_replicates_and_reencrypts("sse-s3", false).await
@@ -4667,7 +4633,6 @@ async fn test_bucket_replication_sse_s3_contract() -> TestResult {
 /// fail closed — replication FAILED, and no plaintext (or any) replica ever
 /// materializes on the target.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_s3_fails_closed_without_target_kms() -> TestResult {
    init_logging();

@@ -4711,7 +4676,6 @@ async fn test_bucket_replication_sse_s3_fails_closed_without_target_kms() -> Tes
 /// the ETag comparison sees the preserved source ETag on the replica and does
 /// not rewrite it, so the replica's version stays stable through the resync.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_s3_resync_converges() -> TestResult {
    init_logging();

@@ -4768,7 +4732,6 @@ async fn test_bucket_replication_sse_s3_resync_converges() -> TestResult {
 /// re-encrypts under its own default key. The independent-KMS pair proves the
 /// replica's envelope is target-owned.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_kms_contract() -> TestResult {
    init_logging();
    assert_managed_sse_replicates_and_reencrypts("sse-kms", true).await
@@ -4779,7 +4742,6 @@ async fn test_bucket_replication_sse_kms_contract() -> TestResult {
 /// carries the full header set (SSE intent, content-type, user metadata) and
 /// the completed replica preserves the source's multipart ETag.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_sse_s3_multipart_reencrypts() -> TestResult {
    init_logging();

@@ -4873,7 +4835,6 @@ async fn test_bucket_replication_sse_s3_multipart_reencrypts() -> TestResult {
 /// still-running source's data scanner (short cycle via [`FAST_SCANNER_ENV`])
 /// re-drives the failed objects once the target is reachable again.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_recovers_after_target_outage() -> TestResult {
    init_logging();

@@ -4953,7 +4914,6 @@ async fn test_bucket_replication_recovers_after_target_outage() -> TestResult {
 /// must settle back to zero even though the historical failed counter remains
 /// non-zero.
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_backlog_metrics_observe_outage_and_recovery() -> TestResult {
    init_logging();

@@ -5087,7 +5047,6 @@ async fn test_bucket_replication_backlog_metrics_observe_outage_and_recovery() -
 /// must converge every persisted failure, including the replayed delete marker
 /// (whose replication decision is re-derived from the live config).
 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_replays_failed_entries_after_source_restart() -> TestResult {
    init_logging();

@@ -5179,7 +5138,6 @@ async fn test_bucket_replication_replays_failed_entries_after_source_restart() -
 }

 #[tokio::test]
-#[serial]
 async fn test_bucket_replication_replayed_delete_marker_preserves_source_mtime_without_source_restart() -> TestResult {
    init_logging();

@@ -5249,7 +5207,6 @@ async fn test_bucket_replication_replayed_delete_marker_preserves_source_mtime_w
 }

 #[tokio::test]
-#[serial]
 async fn test_sequential_bucket_replication_succeeds_for_multiple_buckets() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -5293,7 +5250,6 @@ async fn test_sequential_bucket_replication_succeeds_for_multiple_buckets() -> R
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_recovers_after_runtime_target_cache_is_cleared() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -5337,7 +5293,6 @@ async fn test_replication_recovers_after_runtime_target_cache_is_cleared() -> Re
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_allows_self_signed_https_with_skip_tls_verify_real_dual_node() -> TestResult {
    init_logging();

@@ -5416,7 +5371,6 @@ async fn test_site_replication_allows_self_signed_https_with_skip_tls_verify_rea
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_allows_private_ca_https_with_ca_cert_pem_real_dual_node() -> TestResult {
    init_logging();

@@ -5495,7 +5449,6 @@ async fn test_site_replication_allows_private_ca_https_with_ca_cert_pem_real_dua
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_resync_lifecycle_survives_real_server_restart() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();
    let resync_process_env = [
@@ -5715,7 +5668,6 @@ async fn test_site_replication_resync_lifecycle_survives_real_server_restart() -
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_edit_and_status_peer_state_real_three_node() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -5964,7 +5916,6 @@ async fn test_site_replication_edit_and_status_peer_state_real_three_node() -> R
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_remove_all_real_dual_node() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -6084,7 +6035,6 @@ async fn test_site_replication_remove_all_real_dual_node() -> Result<(), Box<dyn
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_state_edit_fresh_and_stale_real_dual_node() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -6193,7 +6143,6 @@ async fn test_site_replication_state_edit_fresh_and_stale_real_dual_node() -> Re
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_replicates_object_with_bucket_versioning_real_dual_node() -> TestResult {
    init_logging();

@@ -6284,7 +6233,6 @@ async fn test_site_replication_replicates_object_with_bucket_versioning_real_dua
 /// receiver was dropped with only a debug line, while `replicate status` still reported
 /// "1/1 Buckets in sync" because both configs were byte-identical.
 #[tokio::test]
-#[serial]
 async fn test_site_replication_config_broadcast_keeps_reverse_direction_real_dual_node() -> TestResult {
    init_logging();

@@ -6423,7 +6371,6 @@ async fn wait_for_site_replication_rule(
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_active_active_converges_without_loops_real_dual_node() -> TestResult {
    init_logging();

@@ -6741,7 +6688,6 @@ async fn test_site_replication_active_active_converges_without_loops_real_dual_n
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_replicates_policy_backed_user_access_real_dual_node() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -6829,7 +6775,6 @@ async fn test_site_replication_replicates_policy_backed_user_access_real_dual_no
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_replicates_group_policy_backed_access_real_dual_node() -> Result<(), Box<dyn Error + Send + Sync>>
 {
    init_logging();
@@ -6920,7 +6865,6 @@ async fn test_site_replication_replicates_group_policy_backed_access_real_dual_n
 }

 #[tokio::test]
-#[serial]
 async fn test_service_account_policy_from_accountinfo_round_trips_real_single_node() -> TestResult {
    init_logging();

@@ -6972,7 +6916,6 @@ async fn test_service_account_policy_from_accountinfo_round_trips_real_single_no
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_replicates_multiple_service_accounts_real_dual_node() -> Result<(), Box<dyn Error + Send + Sync>> {
    init_logging();

@@ -7073,7 +7016,6 @@ async fn test_site_replication_replicates_multiple_service_accounts_real_dual_no
 }

 #[tokio::test]
-#[serial]
 async fn test_site_replication_replicates_service_accounts_created_from_sts_session_real_dual_node() -> TestResult {
    init_logging();

@@ -7214,7 +7156,6 @@ async fn wait_for_target_request_version_id(
 }

 #[tokio::test]
-#[serial]
 async fn test_bucket_resync_restart_revisits_objects_before_out_of_order_checkpoint() -> TestResult {
    init_logging();

@@ -7333,7 +7274,6 @@ async fn test_bucket_resync_restart_revisits_objects_before_out_of_order_checkpo
 /// CreateMultipartUpload (the version is decided at initiate time) must both
 /// carry the source version as `?versionId=`.
 #[tokio::test]
-#[serial]
 async fn test_replication_put_and_create_multipart_carry_source_version_id_query() -> TestResult {
    init_logging();

@@ -7448,7 +7388,6 @@ async fn test_replication_put_and_create_multipart_carry_source_version_id_query
 /// flow to the onward bucket, proving B's outbound replication and scanner
 /// are live.
 #[tokio::test]
-#[serial]
 async fn test_scanner_never_cascades_inbound_replicas() -> TestResult {
    init_logging();

@@ -7529,7 +7468,6 @@ async fn test_scanner_never_cascades_inbound_replicas() -> TestResult {
 /// version ids and still mint its own there — the check must not report OK
 /// while multipart deletes and heals would silently miss.
 #[tokio::test]
-#[serial]
 async fn test_replication_check_flags_multipart_only_version_minting_target() -> TestResult {
    init_logging();

@@ -7602,7 +7540,6 @@ async fn test_replication_check_flags_multipart_only_version_minting_target() ->
 }

 #[tokio::test]
-#[serial]
 async fn test_replication_check_aborts_failed_multipart_probes() -> TestResult {
    init_logging();

@@ -7765,7 +7702,6 @@ async fn test_replication_check_aborts_failed_multipart_probes() -> TestResult {
 /// BucketRemoteTargetVersionMismatch — while still cleaning up the probe
 /// object via the version id the target actually assigned.
 #[tokio::test]
-#[serial]
 async fn test_replication_check_flags_version_minting_target() -> TestResult {
    init_logging();

@@ -8006,7 +7942,6 @@ async fn wait_for_target_marker_purged(
 /// the target forever. Contract under test: a failed purge attempt is retried
 /// within the watch window and converges once the fault clears.
 #[tokio::test]
-#[serial]
 async fn test_delayed_delete_marker_purge_retries_after_transient_target_failure() -> TestResult {
    init_logging();
    let source_bucket = "delayed-purge-retry-src";
@@ -8066,7 +8001,6 @@ async fn test_delayed_delete_marker_purge_retries_after_transient_target_failure
 /// with an idempotent 204, which used to look like success and strand the
 /// real marker on the target forever.
 #[tokio::test]
-#[serial]
 async fn test_delayed_delete_marker_purge_uses_target_assigned_version() -> TestResult {
    init_logging();
    let source_bucket = "delayed-purge-mint-src";
@@ -8099,7 +8033,6 @@ async fn test_delayed_delete_marker_purge_uses_target_assigned_version() -> Test
 /// replayed purge succeeds, the entry must be acknowledged instead of being
 /// retained as Missed forever.
 #[tokio::test]
-#[serial]
 async fn test_delayed_delete_marker_purge_exhaustion_persists_to_mrf_and_replays_on_restart() -> TestResult {
    init_logging();
    let source_bucket = "delayed-purge-mrf-src";
@@ -8236,7 +8169,6 @@ async fn build_scanner_compensation_pair(
 /// nil-version objects entirely (`scanner_folder.rs` heal_replication), so it
 /// must NEVER be compensated.
 #[tokio::test]
-#[serial]
 async fn test_scanner_compensates_existing_objects_across_write_paths() -> TestResult {
    init_logging();
    let source_bucket = "scanner-comp-src";
@@ -8352,7 +8284,6 @@ async fn test_scanner_compensates_existing_objects_across_write_paths() -> TestR
 /// written after the rule replicate normally (the setting only gates the
 /// existing-object resync path).
 #[tokio::test]
-#[serial]
 async fn test_scanner_never_compensates_when_existing_object_replication_disabled() -> TestResult {
    init_logging();
    let source_bucket = "scanner-disabled-src";
@@ -273,6 +273,7 @@ proptest = "1"
 rcgen.workspace = true
 insta = { workspace = true, features = ["yaml", "json"] }
 rustfs-crypto = { workspace = true }
+tonic-prost = { workspace = true }

 [build-dependencies]
 shadow-rs = { workspace = true, default-features = false, features = ["build", "metadata"] }
@@ -373,14 +373,14 @@ pub mod error {

 pub mod erasure {
    pub use crate::erasure::coding::{
-        BitrotReader, BitrotWriter, BitrotWriterWrapper, CustomWriter, Erasure, ErasureConstructionError, ReedSolomonEncoder,
-        calc_shard_size, calc_shard_size_legacy,
+        BitrotReader, BitrotSelfTestError, BitrotWriter, BitrotWriterWrapper, CustomWriter, Erasure, ErasureConstructionError,
+        ReedSolomonEncoder, bitrot_self_test, calc_shard_size, calc_shard_size_legacy,
    };
 }

 pub mod event {
    pub use crate::event::name::EventName;
-    pub use crate::services::event_notification::{EventArgs, register_event_dispatch_hook};
+    pub use crate::services::event_notification::{EventArgs, register_event_dispatch_hook, send_event};
 }

 pub mod global {
@@ -483,6 +483,7 @@ pub mod store_list {
 }

 pub mod storage {
+    pub use crate::core::pools::HealLifecycleExpiryContext;
    pub use crate::store::HealWalkVersion;
    pub use crate::store::{
        ECStore, all_local_disk, all_local_disk_path, find_local_disk_by_ref, init_local_disks,
@@ -82,7 +82,6 @@ use tracing::warn;
 use url::Url;
 use uuid::Uuid;

-const DEFAULT_HEALTH_CHECK_RELOAD_DURATION: Duration = Duration::from_secs(30 * 60);
 const MAX_CONCURRENT_TARGET_HEALTH_CHECKS: usize = 16;
 const REDACTED_CREDENTIAL: &str = "<redacted>";

@@ -1550,8 +1549,8 @@ impl Default for PutObjectOptions {
    }
 }

-#[allow(dead_code)]
 impl PutObjectOptions {
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn set_match_etag(&mut self, etag: &str) {
        if etag == "*" {
            self.custom_header
@@ -1562,6 +1561,7 @@ impl PutObjectOptions {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn set_match_etag_except(&mut self, etag: &str) {
        if etag == "*" {
            self.custom_header
@@ -1697,6 +1697,7 @@ impl PutObjectOptions {
        header
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn validate(&self, _c: Arc<TargetClient>) -> Result<(), std::io::Error> {
        //if self.checksum.is_set() {
        /*if !self.trailing_header_support {
@@ -126,11 +126,23 @@ const EVENT_LIFECYCLE_EXPIRED_DETECTED: &str = "lifecycle_expired_detected";
 const EVENT_LIFECYCLE_NOT_ENQUEUED: &str = "lifecycle_not_enqueued";
 const EVENT_LIFECYCLE_DELETE_DISPATCHED: &str = "lifecycle_delete_dispatched";
 const EVENT_LIFECYCLE_DELETE_COMPLETED: &str = "lifecycle_delete_completed";
+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 const EVENT_LIFECYCLE_TIER_AUDIT: &str = "lifecycle_tier_audit";
 const EVENT_LIFECYCLE_TIER_OPERATION_FAILED: &str = "lifecycle_tier_operation_failed";
 const EVENT_LIFECYCLE_DELETE_FAILED: &str = "lifecycle_delete_failed";

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub type TimeFn = Arc<dyn Fn() -> Pin<Box<dyn Future<Output = ()> + Send>> + Send + Sync + 'static>;
+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub type TraceFn =
    Arc<dyn Fn(String, HashMap<String, String>) -> Pin<Box<dyn Future<Output = ()> + Send>> + Send + Sync + 'static>;
 pub type ExpiryOpType = Box<dyn ExpiryOp + Send + Sync + 'static>;
@@ -140,9 +152,21 @@ static TIER_FREE_VERSION_RECOVERY_STARTED: OnceLock<()> = OnceLock::new();
 static MANUAL_TRANSITION_JOB_RECOVERY_STARTED: OnceLock<()> = OnceLock::new();

 pub const AMZ_OBJECT_TAGGING: &str = "X-Amz-Tagging";
+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub const AMZ_TAG_COUNT: &str = "x-amz-tagging-count";
+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub const AMZ_TAG_DIRECTIVE: &str = "X-Amz-Tagging-Directive";
 pub const AMZ_ENCRYPTION_AES: &str = "AES256";
+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub const AMZ_ENCRYPTION_KMS: &str = "aws:kms";

 pub const ERR_INVALID_STORAGECLASS: &str = "invalid tier.";
@@ -280,6 +304,10 @@ impl LifecycleSys {
        }
    }

+    #[allow(
+        dead_code,
+        reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+    )]
    pub fn trace(oi: &ObjectInfo) -> TraceFn {
        let bucket = oi.bucket.clone();
        let name = oi.name.clone();
@@ -428,16 +456,23 @@ impl<'a> LifecycleExpiryTrace<'a> {
    }
 }

-#[allow(dead_code)]
 impl ExpiryStats {
    pub fn missed_tasks(&self) -> i64 {
        self.missed_expiry_tasks.load(Ordering::SeqCst)
    }

+    #[allow(
+        dead_code,
+        reason = "asserted by this file's tests; the lib target cannot see test-only consumers (backlog#1823)"
+    )]
    fn missed_free_vers_tasks(&self) -> i64 {
        self.missed_freevers_tasks.load(Ordering::SeqCst)
    }

+    #[allow(
+        dead_code,
+        reason = "asserted by this file's tests; the lib target cannot see test-only consumers (backlog#1823)"
+    )]
    fn missed_tier_journal_tasks(&self) -> i64 {
        self.missed_tier_journal_tasks.load(Ordering::SeqCst)
    }
@@ -570,6 +605,10 @@ async fn delete_free_version_remote_object(
    Ok(())
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 async fn delete_free_version_remote_object_then<T, F, Fut>(
    oi: &ObjectInfo,
    tier_config_mgr: &Arc<RwLock<TierConfigMgr>>,
@@ -1744,7 +1783,7 @@ impl TransitionState {
                                .await;
                            }
                            global_metrics().record_scanner_transition_failed(1);
-                            if !is_err_version_not_found(&err) && !is_err_object_not_found(&err) && !is_network_or_host_down(&err.to_string(), false) && !err.to_string().contains("use of closed network connection") {
+                            if !is_err_version_not_found(&err) && !is_err_object_not_found(&err) && !is_network_or_host_down(&err.to_string(), false) {
                                error!(
                                    event = EVENT_LIFECYCLE_TIER_OPERATION_FAILED,
                                    component = LOG_COMPONENT_ECSTORE,
@@ -2868,6 +2907,10 @@ fn stale_upload_default_due(initiated: OffsetDateTime, default_expiry: StdDurati
    initiated + time::Duration::seconds(default_expiry.as_secs() as i64)
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 async fn stale_upload_current_size(set: &Arc<SetDisks>, metadata: &HashMap<String, String>, upload_dir: &str) -> Option<usize> {
    stale_upload_current_size_with_opts(set, metadata, upload_dir, false).await
 }
@@ -3352,6 +3395,10 @@ pub async fn validate_transition_tier(lc: &BucketLifecycleConfiguration) -> Resu
    Ok(())
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 fn mark_delete_opts_skip_decommissioned_on_remote_success(opts: &mut ObjectOptions, remote_delete_succeeded: bool) {
    if remote_delete_succeeded {
        opts.skip_decommissioned = true;
@@ -4339,6 +4386,10 @@ pub async fn expire_transitioned_object(
    Ok(dobj)
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub fn gen_transition_objname(bucket: &str) -> Result<String, Error> {
    let us = Uuid::new_v4().to_string();
    let mut hasher = Sha256::new();
@@ -4373,6 +4424,10 @@ pub async fn transition_object(api: Arc<ECStore>, oi: &ObjectInfo, lae: LcAuditE
    result
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub fn audit_tier_actions(_tier: &str, bytes: i64) -> TimeFn {
    let tier = _tier.to_string();
    Arc::new(move || {
@@ -4391,6 +4446,10 @@ pub fn audit_tier_actions(_tier: &str, bytes: i64) -> TimeFn {
    })
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn get_transitioned_object_reader(
    bucket: &str,
    object: &str,
@@ -5145,6 +5204,10 @@ async fn lifecycle_delete_config_snapshot(api: &ECStore, oi: &ObjectInfo) -> Res
    ReplicationObjectBridge::delete_request_config(api, &oi.bucket).await
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn apply_lifecycle_action(event: &lifecycle::Event, src: &LcEventSrc, oi: &ObjectInfo) -> bool {
    let mut success = false;
    match event.action {
@@ -7422,6 +7485,10 @@ mod tests {
    // process environment while `env::set_var`/`env::remove_var` is active.
    // SAFETY: keep this note adjacent to the allowance for the repository guard.
    #[allow(unsafe_code)]
+    #[allow(
+        dead_code,
+        reason = "transition-queue env fixture kept for tests that scope those vars; no test uses it today (backlog#1823)"
+    )]
    async fn with_transition_queue_env_async<F, Fut>(capacity: Option<&str>, timeout_ms: Option<&str>, test_fn: F)
    where
        F: FnOnce() -> Fut,
@@ -759,6 +759,10 @@ pub struct ManualTransitionWorkerResultRecord {
 }

 impl ManualTransitionWorkerResultRecord {
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+    )]
    pub fn new(job_id: Uuid, task_key: impl Into<String>, result: ManualTransitionWorkerResult) -> Self {
        Self::new_with_reason(job_id, task_key, result, None)
    }
@@ -1257,6 +1261,10 @@ pub(crate) async fn save_manual_transition_task_if_absent(
    }
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn load_manual_transition_task_record(
    api: Arc<ECStore>,
    job_id: Uuid,
@@ -1320,6 +1328,10 @@ async fn scan_manual_transition_task_journal(api: Arc<ECStore>, job_id: Uuid) ->
    }
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn load_manual_transition_worker_result_stats(
    api: Arc<ECStore>,
    job_id: Uuid,
@@ -1455,6 +1467,10 @@ async fn scan_manual_transition_worker_result_journal(
    }
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn reconcile_manual_transition_worker_results(
    api: Arc<ECStore>,
    job_id: Uuid,
@@ -19,7 +19,7 @@ pub mod core;
 pub mod evaluator;
 pub mod manual_transition_job;
 mod metadata_boundary;
-pub(crate) use metadata_boundary::get_expiry_configs;
+pub(crate) use metadata_boundary::{LifecycleExpiryConfigs, get_expiry_configs};
 mod object_lock_boundary;
 pub use self::core as lifecycle;
 mod replication_sink;
@@ -15,25 +15,35 @@
 use rustfs_common::metrics::IlmAction;

 use crate::bucket::lifecycle::lifecycle::ObjectOpts;
+use crate::bucket::replication::ReplicationLifecycleBridge;
 pub(crate) use crate::bucket::replication::ReplicationStatusType;
 #[cfg(test)]
 pub(crate) use crate::bucket::replication::VersionPurgeStatusType;
 pub(crate) use crate::bucket::replication::{
    DeleteReplicationConfigSnapshot, ReplicationObjectBridge, replication_state_to_filemeta,
 };
-use crate::bucket::replication::{ReplicationLifecycleBridge, ReplicationLifecycleConfig};
 use crate::storage_api_contracts::object::DeletedObject;

-pub(crate) type LifecycleReplicationConfig = ReplicationLifecycleConfig;
-
+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) fn has_pending_version_purge(obj: &ObjectOpts) -> bool {
    obj.version_purge_status.is_pending()
 }

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) fn has_pending_object_replication(obj: &ObjectOpts) -> bool {
    replication_status_blocks_lifecycle(&obj.replication_status)
 }

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) fn has_pending_lifecycle_replication(obj: &ObjectOpts) -> bool {
    has_pending_object_replication(obj) || has_pending_version_purge(obj)
 }
@@ -14,6 +14,10 @@

 use std::collections::HashMap;

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) fn decode_tags_to_map(tags: &str) -> HashMap<String, String> {
    crate::bucket::tagging::decode_tags_to_map(tags)
 }
@@ -331,6 +331,10 @@ where
    persist_tier_delete_journal_entry(api, &committed).await
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn abort_tier_delete_journal_entry<S>(api: Arc<S>, je: &Jentry) -> std::io::Result<()>
 where
    S: ObjectOperations<
@@ -148,6 +148,10 @@ struct RecoveryCursor {
    object: String,
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn recover_tier_free_versions(
    api: Arc<ECStore>,
    limit: usize,
@@ -80,7 +80,10 @@ impl LastDayTierStats {
        }
    }

-    #[allow(dead_code)]
+    #[allow(
+        dead_code,
+        reason = "asserted by this file's tests; the lib target cannot see test-only consumers (backlog#1823)"
+    )]
    fn merge(&self, m: LastDayTierStats) -> LastDayTierStats {
        let mut cl = self.clone();
        let mut cm = m;
@@ -177,9 +177,10 @@ fn should_record_remote_delete_failure(err: &std::io::Error) -> bool {
 }

 #[derive(Default)]
-#[allow(dead_code)]
 struct ObjSweeper {
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    object: String,
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    bucket: String,
    version_id: Option<Uuid>,
    versioned: bool,
@@ -191,9 +192,9 @@ struct ObjSweeper {
    remote_object: String,
 }

-#[allow(dead_code)]
 impl ObjSweeper {
    #[allow(clippy::new_ret_no_self)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub async fn new(bucket: &str, object: &str) -> Result<Self, std::io::Error> {
        Ok(Self {
            object: object.into(),
@@ -202,17 +203,20 @@ impl ObjSweeper {
        })
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub fn with_version(&mut self, vid: Option<Uuid>) -> &Self {
        self.version_id = vid.clone();
        self
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub fn with_versioning(&mut self, versioned: bool, suspended: bool) -> &Self {
        self.versioned = versioned;
        self.suspended = suspended;
        self
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub fn get_opts(&self) -> lifecycle::ObjectOpts {
        let mut opts = ObjectOpts {
            version_id: self.version_id.clone(),
@@ -226,6 +230,7 @@ impl ObjSweeper {
        opts
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub fn set_transition_state(&mut self, info: TransitionedObject) {
        self.transition_tier = info.tier;
        self.transition_status = info.status;
@@ -266,6 +271,7 @@ impl ObjSweeper {
        None
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    pub async fn sweep(&self, api: Arc<ECStore>) {
        let Some(je) = self.should_remove_remote_object() else {
            return;
@@ -385,6 +391,10 @@ impl ExpiryOp for Jentry {
    }
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn delete_object_from_remote_tier(obj_name: &str, rv_id: &str, tier_name: &str) -> Result<(), std::io::Error> {
    let result = delete_object_from_remote_tier_raw(obj_name, rv_id, tier_name).await;
    if let Err(err) = &result
@@ -395,6 +405,10 @@ pub async fn delete_object_from_remote_tier(obj_name: &str, rv_id: &str, tier_na
    result
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 async fn delete_object_from_remote_tier_raw(obj_name: &str, rv_id: &str, tier_name: &str) -> Result<(), std::io::Error> {
    #[cfg(test)]
    if let Some(result) = run_remote_tier_delete_test_hook(obj_name, rv_id, tier_name) {
@@ -405,6 +419,10 @@ async fn delete_object_from_remote_tier_raw(obj_name: &str, rv_id: &str, tier_na
    delete_object_from_remote_tier_raw_with_manager(obj_name, rv_id, tier_name, &tier_config_mgr).await
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 async fn delete_object_from_remote_tier_raw_with_manager(
    obj_name: &str,
    rv_id: &str,
@@ -485,6 +503,10 @@ pub enum RemoteTierDeleteOutcome {
    AlreadyRemoved,
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+)]
 pub async fn delete_object_from_remote_tier_idempotent(
    obj_name: &str,
    rv_id: &str,
@@ -50,8 +50,16 @@ pub type Result<T> = std::result::Result<T, TransitionTransactionError>;
 #[derive(Debug, thiserror::Error)]
 pub enum TransitionTransactionError {
    #[error("transition transaction already exists")]
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+    )]
    AlreadyExists,
    #[error("transition transaction is not found")]
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity tier/lifecycle entry point that this port never wired (backlog#1823)"
+    )]
    NotFound,
    #[error("transition transaction is corrupt: {0}")]
    Corrupt(&'static str),
@@ -60,12 +60,14 @@ struct ConfigWriteLockProbeState {
 static CONFIG_WRITE_LOCK_PROBES: std::sync::OnceLock<StdMutex<Vec<Arc<ConfigWriteLockProbeState>>>> = std::sync::OnceLock::new();

 #[cfg(any(test, feature = "test-util"))]
+#[allow(dead_code, reason = "installed by tests behind `--features test-util` (backlog#1823)")]
 pub struct ConfigWriteLockProbe {
    state: Arc<ConfigWriteLockProbeState>,
 }

 #[cfg(any(test, feature = "test-util"))]
 impl ConfigWriteLockProbe {
+    #[allow(dead_code, reason = "installed by tests behind `--features test-util` (backlog#1823)")]
    pub fn install(bucket: &str) -> Self {
        let state = Arc::new(ConfigWriteLockProbeState {
            bucket: bucket.to_string(),
@@ -84,6 +86,7 @@ impl ConfigWriteLockProbe {
        Self { state }
    }

+    #[allow(dead_code, reason = "installed by tests behind `--features test-util` (backlog#1823)")]
    pub async fn wait_until_attempted(&self) {
        tokio::time::timeout(Duration::from_secs(30), self.state.arrived.notified())
            .await
@@ -890,6 +893,10 @@ pub async fn get_bucket_policy_raw(bucket: &str) -> Result<(String, OffsetDateTi
    bucket_meta_sys.get_bucket_policy_raw(bucket).await
 }

+#[allow(
+    dead_code,
+    reason = "free-function facade over the live BucketMetadataSys::get_bucket_acl_config; no caller in this port (backlog#1823)"
+)]
 pub async fn get_bucket_acl_config(bucket: &str) -> Result<(String, OffsetDateTime)> {
    let bucket_meta_sys_lock = get_bucket_metadata_sys()?;
    let bucket_meta_sys = bucket_meta_sys_lock.read().await;
@@ -1104,6 +1111,10 @@ pub async fn get_config_from_disk(bucket: &str) -> Result<BucketMetadata> {
    bucket_meta_sys.get_config_from_disk(bucket).await
 }

+#[allow(
+    dead_code,
+    reason = "ambient-facade variant of the live created_at_in; no caller in this port (backlog#1823)"
+)]
 pub async fn created_at(bucket: &str) -> Result<OffsetDateTime> {
    let bucket_meta_sys_lock = get_bucket_metadata_sys()?;
    let bucket_meta_sys = bucket_meta_sys_lock.read().await;
@@ -1617,6 +1628,7 @@ impl BucketMetadataSys {
    /// [`Self::update`], with the payload computed from the loaded metadata
    /// instead of supplied up front. Loads through this system's own store so
    /// the read and the persisted write target the same instance.
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    async fn update_config_with<F>(&self, bucket: &str, config_file: &str, mutate: F) -> Result<OffsetDateTime>
    where
        F: FnOnce(&BucketMetadata) -> Result<Vec<u8>> + Send,
@@ -1721,6 +1733,7 @@ impl BucketMetadataSys {
    /// A miss is never published as an authoritative default, and a snapshot
    /// read before delete plus same-name recreation cannot replace the new
    /// generation.
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(crate) async fn reload_from_store(&self, bucket: &str) -> Result<()> {
        if is_meta_bucketname(bucket) {
            return Err(Error::other("errInvalidArgument"));
@@ -13,7 +13,6 @@
 // limitations under the License.

 // #730: bucket subsystems still contain staged ECStore migration code.
-#![allow(dead_code)]

 pub mod bandwidth;
 pub mod bucket_target_sys;
@@ -136,6 +136,7 @@ pub fn add_years(dt: OffsetDateTime, years: i32) -> OffsetDateTime {

 /// Check if an object has legal hold enabled.
 /// Returns true if legal hold is ON.
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn has_legal_hold(user_defined: &std::collections::HashMap<String, String>) -> bool {
    let lhold = objectlock::get_object_legalhold_meta(user_defined);
    matches!(lhold.status, Some(ref st) if st.as_str() == ObjectLockLegalHoldStatus::ON)
@@ -151,6 +152,7 @@ fn has_legal_hold(user_defined: &std::collections::HashMap<String, String>) -> b
 /// # Returns
 /// * `true` if the object is locked (cannot be deleted/modified)
 /// * `false` if the object is not locked
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn is_object_locked_by_metadata(user_defined: &std::collections::HashMap<String, String>, is_delete_marker: bool) -> bool {
    // Delete markers are never locked
    if is_delete_marker {
@@ -193,6 +193,7 @@ pub enum QuotaError {
 }

 #[derive(Debug, Serialize)]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub struct QuotaErrorResponse {
    #[serde(rename = "Code")]
    pub code: String,
@@ -208,6 +209,7 @@ pub struct QuotaErrorResponse {
 }

 impl QuotaErrorResponse {
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub fn new(quota_error: &QuotaError, request_id: &str, host_id: &str) -> Self {
        match quota_error {
            QuotaError::QuotaExceeded { .. } => Self {
@@ -310,9 +312,7 @@ mod tests {
        }
        #[derive(Deserialize)]
        struct LegacyBucketQuota {
-            #[allow(dead_code)]
            quota: Option<u64>,
-            #[allow(dead_code)]
            quota_type: LegacyQuotaType,
        }
        let legacy = serde_json::from_slice::<LegacyBucketQuota>(&json)
@@ -899,6 +899,7 @@ async fn save_ledger_locked(
 }

 #[cfg(any(test, feature = "test-util"))]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn fail_next_quota_ledger_save_for_test() {
    FAIL_NEXT_LEDGER_SAVE.store(true, std::sync::atomic::Ordering::SeqCst);
 }
@@ -60,7 +60,7 @@ pub use replication_filemeta_boundary::{
 pub(crate) use replication_filemeta_boundary::{
    replication_state_from_filemeta, replication_status_from_filemeta, version_purge_status_from_filemeta,
 };
-pub(crate) use replication_lifecycle_bridge::{ReplicationLifecycleBridge, ReplicationLifecycleConfig};
+pub(crate) use replication_lifecycle_bridge::ReplicationLifecycleBridge;
 pub(crate) use replication_migration_bridge::ReplicationMigrationBridge;
 pub use replication_object_bridge::ReplicationObjectBridge;
 pub use replication_object_config::{DeleteReplicationConfigSnapshot, ReplicationConfig};
@@ -37,6 +37,10 @@ impl ReplicationConfigStore {
        com::read_config_limited(api, file, max_bytes).await
    }

+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    pub(crate) async fn read_no_lock<S>(api: Arc<S>, file: &str) -> Result<Vec<u8>>
    where
        S: ReplicationObjectIO,
@@ -24,15 +24,27 @@ use super::replication_storage_boundary::{
    DeletedObject, ObjectInfo, ObjectOptions, ObjectToDelete, deleted_object_for_replication,
 };

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) type ReplicationLifecycleConfig = ReplicationConfig;

 pub(crate) struct ReplicationLifecycleBridge;

 impl ReplicationLifecycleBridge {
+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn new_config(config: ReplicationConfiguration) -> ReplicationLifecycleConfig {
        ReplicationConfig::new(Some(config), None)
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn has_pending_version_purge(
        config: &ReplicationLifecycleConfig,
        object_name: &str,
@@ -45,6 +57,10 @@ impl ReplicationLifecycleBridge {
                .is_some_and(|config| config.has_active_rules(object_name, true))
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) async fn check_delete_replication(
        bucket: &str,
        object: &ObjectToDelete,
@@ -54,6 +70,10 @@ impl ReplicationLifecycleBridge {
        check_replicate_delete(bucket, object, source, opts, None).await
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn version_delete_replication_state(decision: &ReplicateDecision) -> ReplicationState {
        let pending_status = decision.pending_status();
        ReplicationState {
@@ -19,17 +19,33 @@ use time::OffsetDateTime;
 use super::replication_error_boundary::Result;
 use crate::bucket::msgp_decode;

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) struct ReplicationMsgpCodec;

 impl ReplicationMsgpCodec {
+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn read_ext8_time<R: Read>(rd: &mut R) -> Result<OffsetDateTime> {
        msgp_decode::read_msgp_ext8_time(rd)
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn skip_value<R: Read>(rd: &mut R) -> Result<()> {
        msgp_decode::skip_msgp_value(rd)
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn write_time<W: Write>(wr: &mut W, time: OffsetDateTime) -> Result<()> {
        msgp_decode::write_msgp_time(wr, time)
    }
@@ -77,6 +77,10 @@ impl ReplicationObjectBridge {
        load_delete_request_config_in(ctx, bucket).await
    }

+    #[allow(
+        dead_code,
+        reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+    )]
    pub(crate) async fn delete_config_snapshot_in(
        ctx: &ReplicationInstanceContext,
        bucket: &str,
@@ -231,6 +231,10 @@ pub(crate) async fn load_delete_replication_config(
    delete_snapshot_from_metadata(ReplicationMetadataStore::delete_metadata(bucket).await?)
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+)]
 pub(crate) async fn load_delete_replication_config_in(
    ctx: &ReplicationInstanceContext,
    bucket: &str,
@@ -217,6 +217,10 @@ impl DurableMrfBacklogTracker {
    }
 }

+#[allow(
+    dead_code,
+    reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+)]
 fn durable_mrf_backlog_tracker_from_entries(entries: &[MrfReplicateEntry]) -> DurableMrfBacklogTracker {
    let mut tracker = DurableMrfBacklogTracker {
        available: true,
@@ -712,6 +716,10 @@ pub struct ReplicationPool<S: ReplicationStorage> {

    // MRF worker lifecycle
    mrf_worker_cancellations: Mutex<Vec<CancellationToken>>,
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    mrf_stop_tx: Sender<()>,

    // Worker size tracking
@@ -940,6 +948,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Resizes worker priority and counts
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    pub async fn resize_worker_priority(
        &self,
        pri: ReplicationPriority,
@@ -1180,6 +1192,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Queues an MRF save operation
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    async fn queue_mrf_save(&self, entry: MrfReplicateEntry) {
        let _ = self.queue_mrf_save_admission(entry, "mrf_worker").await;
    }
@@ -1651,6 +1667,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Worker function for handling regular replication operations
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    async fn add_worker(
        &self,
        mut rx: Receiver<ReplicationOperation>,
@@ -1664,6 +1684,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Worker function for handling large object replication operations
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    async fn add_large_worker(
        &self,
        mut rx: Receiver<ReplicationOperation>,
@@ -1678,6 +1702,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Worker function for handling MRF (Most Recent Failures) operations
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    async fn add_mrf_worker(
        &self,
        mut rx: Receiver<ReplicationOperation>,
@@ -1691,6 +1719,10 @@ impl<S: ReplicationStorage> ReplicationPool<S> {
    }

    /// Delete resync metadata from replication resync state in memory
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    pub async fn delete_resync_metadata(&self, bucket: &str) {
        let mut status_map = self.resyncer.status_map.write().await;
        status_map.remove(bucket);
@@ -21,11 +21,31 @@ pub(crate) use rustfs_replication::{
    should_count_head_proxy_failure,
 };

+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) const RESYNC_META_FORMAT: u16 = rustfs_replication::resync::RESYNC_META_FORMAT;
+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) const RESYNC_META_VERSION: u16 = rustfs_replication::resync::RESYNC_META_VERSION;
 pub(crate) const RESYNC_FILE_MAX_BYTES: usize = rustfs_replication::RESYNC_FILE_MAX_BYTES;
+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) const WIRE_ZERO_TIME_UNIX: i64 = rustfs_replication::resync::WIRE_ZERO_TIME_UNIX;
+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) const MRF_META_FORMAT: u16 = rustfs_replication::mrf::MRF_META_FORMAT;
+#[allow(
+    dead_code,
+    reason = "declared boundary surface for the ECStore replication split plan; no caller in this port (backlog#1823)"
+)]
 pub(crate) const MRF_META_VERSION: u16 = rustfs_replication::mrf::MRF_META_VERSION;

 fn map_replication_error(err: rustfs_replication::Error) -> Error {
@@ -122,6 +122,10 @@ const REPLICATION_TARGET_OFFLINE_ERROR_MARKERS: &[&str] = &[
    "tcp connect error",
 ];

+#[allow(
+    dead_code,
+    reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+)]
 const RESYNC_TIME_INTERVAL: TokioDuration = TokioDuration::from_secs(60);

 static WARNED_MONITOR_UNINIT: std::sync::Once = std::sync::Once::new();
@@ -328,6 +332,10 @@ fn bounded_resync_max_jobs(value: usize) -> usize {
 #[derive(Debug)]
 pub struct ReplicationResyncer {
    pub status_map: Arc<RwLock<HashMap<String, BucketReplicationResyncStatus>>>,
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    pub worker_size: usize,
    pub(crate) cancel_tokens: Arc<RwLock<HashMap<ResyncCancelKey, CancellationToken>>>,
    resync_admission: Arc<Semaphore>,
@@ -544,6 +552,10 @@ impl ReplicationResyncer {
            .is_some_and(|status| status.failed_count > 0)
    }

+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    pub async fn persist_to_disk<S>(&self, cancel_token: CancellationToken, api: Arc<S>)
    where
        S: ReplicationObjectIO,
@@ -340,6 +340,10 @@ impl ReplicationStats {
    }

    /// Site replication update replica statistics
+    #[allow(
+        dead_code,
+        reason = "MinIO-parity replication surface with no caller in this port (backlog#1823)"
+    )]
    fn sr_update_replica_stat(&self, size: i64) {
        self.sr_stats.replica_size.fetch_add(size, Ordering::Relaxed);
        self.sr_stats.replica_count.fetch_add(1, Ordering::Relaxed);
@@ -59,6 +59,10 @@ impl fmt::Debug for Credentials {
 }

 #[derive(Debug, Deserialize, Serialize, Default, Clone)]
+#[allow(
+    dead_code,
+    reason = "MinIO-parity bucket-target service discriminator with no caller in this port (backlog#1823)"
+)]
 pub enum ServiceType {
    #[default]
    Replication,
@@ -73,23 +73,6 @@ pub fn check_valid_bucket_name_strict(bucket_name: &str) -> Result<()> {
    check_bucket_name_common(bucket_name, true)
 }

-pub fn check_valid_object_name_prefix(object_name: &str) -> Result<()> {
-    if object_name.len() > 1024 {
-        return Err(Error::other("Object name cannot be longer than 1024 characters"));
-    }
-    if !object_name.is_ascii() {
-        return Err(Error::other("Object name with non-UTF-8 strings are not supported"));
-    }
-    Ok(())
-}
-
-pub fn check_valid_object_name(object_name: &str) -> Result<()> {
-    if object_name.trim().is_empty() {
-        return Err(Error::other("Object name cannot be empty"));
-    }
-    check_valid_object_name_prefix(object_name)
-}
-
 pub fn deserialize<T>(input: &[u8]) -> xml::DeResult<T>
 where
    T: for<'xml> xml::Deserialize<'xml>,
@@ -100,6 +83,10 @@ where
    Ok(ans)
 }

+#[allow(
+    dead_code,
+    reason = "xml serialize helper with no caller in this port; the live sibling is deserialize (backlog#1823)"
+)]
 pub fn serialize_content<T: xml::SerializeContent>(val: &T) -> xml::SerResult<String> {
    let mut buf = Vec::with_capacity(256);
    {
@@ -186,15 +173,27 @@ pub fn is_valid_object_name(object: &str) -> bool {
 /// Client-facing reason attached to rejections of object keys that Win32/NTFS
 /// cannot represent as file paths (issue #3299). Deployments on Linux/macOS
 /// accept the full S3 key character set.
+#[allow(
+    dead_code,
+    reason = "live on Windows: callers sit inside the #[cfg(target_os = \"windows\")] block in check_object_name_for_length_and_slash (backlog#1823)"
+)]
 pub const WINDOWS_RESERVED_CHARACTERS_REASON: &str =
    "object key contains characters unsupported on Windows hosts (one of ':', '*', '?', '\"', '|', '<', '>')";

 /// Client-facing reason for path segments Windows can store but not address
 /// afterwards (issue #3449): trailing dot/space or reserved DOS device names.
+#[allow(
+    dead_code,
+    reason = "live on Windows: callers sit inside the #[cfg(target_os = \"windows\")] block in check_object_name_for_length_and_slash (backlog#1823)"
+)]
 pub const WINDOWS_RESERVED_SEGMENT_REASON: &str = "object key contains a path segment unsupported on Windows hosts (trailing dot or space, or a reserved device name such as NUL/CON/COM1)";

 /// Reserved DOS device names that shadow regular files on Windows, even when
 /// an extension is appended (e.g. `NUL.txt` resolves to the `NUL` device).
+#[allow(
+    dead_code,
+    reason = "live on Windows: callers sit inside the #[cfg(target_os = \"windows\")] block in check_object_name_for_length_and_slash (backlog#1823)"
+)]
 const WINDOWS_RESERVED_NAMES: &[&str] = &[
    "CON", "PRN", "AUX", "NUL", "COM1", "COM2", "COM3", "COM4", "COM5", "COM6", "COM7", "COM8", "COM9", "LPT1", "LPT2", "LPT3",
    "LPT4", "LPT5", "LPT6", "LPT7", "LPT8", "LPT9",
@@ -204,6 +203,10 @@ const WINDOWS_RESERVED_NAMES: &[&str] = &[
 /// the Win32 API cannot address afterwards (issue #3449): segments ending in a
 /// dot or a space, and reserved DOS device names — bare or with an extension
 /// (`NUL.txt`), matching classic Win32 path resolution semantics.
+#[allow(
+    dead_code,
+    reason = "live on Windows: callers sit inside the #[cfg(target_os = \"windows\")] block in check_object_name_for_length_and_slash (backlog#1823)"
+)]
 pub fn object_name_has_windows_incompatible_segment(object: &str) -> bool {
    object.split(['/', '\\']).any(|segment| {
        if segment.ends_with('.') || segment.ends_with(' ') {
@@ -90,6 +90,10 @@ impl BucketVersioningSys {
    /// caller's own instance context so a second in-process store never
    /// answers with the first instance's versioning state; falls back to the
    /// ambient system when the instance cell is not initialized.
+    #[allow(
+        dead_code,
+        reason = "instance-scoped seam (backlog#1052) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) async fn get_in(ctx: &crate::runtime::instance::InstanceContext, bucket: &str) -> Result<VersioningConfiguration> {
        if bucket == RUSTFS_META_BUCKET || bucket.starts_with(RUSTFS_META_BUCKET) {
            return Ok(VersioningConfiguration::default());
@@ -95,7 +95,6 @@ impl TransitionClient {
 }

 #[derive(Default)]
-#[allow(dead_code)]
 pub struct GetRequest {
    pub buffer: Vec<u8>,
    pub offset: i64,
@@ -107,11 +106,12 @@ pub struct GetRequest {
    pub setting_object_info: bool,
 }

-#[allow(dead_code)]
 pub struct GetResponse {
    pub size: i64,
    //pub error:       error,
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    pub did_read: bool,
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    pub object_info: ObjectInfo,
 }

@@ -20,6 +20,7 @@
 #![allow(clippy::all)]

 use http::{HeaderMap, HeaderName, HeaderValue};
+use rustfs_utils::http::headers::AMZ_CHECKSUM_MODE;
 use std::collections::HashMap;
 use time::OffsetDateTime;
 use tracing::warn;
@@ -27,7 +28,6 @@ use tracing::warn;
 use crate::client::api_error_response::err_invalid_argument;

 #[derive(Default)]
-#[allow(dead_code)]
 pub struct AdvancedGetOptions {
    pub replication_delete_marker: bool,
    pub is_replication_ready_for_delete_marker: bool,
@@ -77,7 +77,7 @@ impl GetObjectOptions {
            }
        }
        if self.checksum {
-            headers.insert(HeaderName::from_static("x-amz-checksum-mode"), HeaderValue::from_static("ENABLED"));
+            headers.insert(HeaderName::from_static(AMZ_CHECKSUM_MODE), HeaderValue::from_static("ENABLED"));
        }
        headers
    }
@@ -360,7 +360,6 @@ impl TransitionClient {
 }

 #[derive(Default)]
-#[allow(dead_code)]
 pub struct ListObjectsOptions {
    reverse_versions: bool,
    with_versions: bool,
@@ -137,8 +137,8 @@ impl Default for PutObjectOptions {
    }
 }

-#[allow(dead_code)]
 impl PutObjectOptions {
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn set_match_etag(&mut self, etag: &str) {
        if etag == "*" {
            self.custom_header.insert("If-Match", HeaderValue::from_static("*"));
@@ -149,6 +149,7 @@ impl PutObjectOptions {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn set_match_etag_except(&mut self, etag: &str) {
        if etag == "*" {
            self.custom_header.insert("If-None-Match", HeaderValue::from_static("*"));
@@ -259,6 +260,7 @@ impl PutObjectOptions {
        header
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn validate(&self, c: TransitionClient) -> Result<(), std::io::Error> {
        //if self.checksum.is_set() {
        /*if !self.trailing_header_support {
@@ -55,7 +55,6 @@ pub struct RemoveBucketOptions {
 const DELETE_RESPONSE_PREVIEW_LEN: usize = 1024;

 #[derive(Debug)]
-#[allow(dead_code)]
 pub struct AdvancedRemoveOptions {
    pub replication_delete_marker: bool,
    pub replication_status: ReplicationStatus,
@@ -465,10 +464,10 @@ impl TransitionClient {
 }

 #[derive(Debug, Default)]
-#[allow(dead_code)]
 pub struct RemoveObjectError {
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    object_name: String,
-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
    version_id: String,
    err: Option<std::io::Error>,
 }
@@ -372,8 +372,8 @@ pub struct Checksum {
    computed: bool,
 }

-#[allow(dead_code)]
 impl Checksum {
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn new(t: ChecksumMode, b: &[u8]) -> Checksum {
        if t.is_set() && b.len() == t.raw_byte_len() {
            return Checksum {
@@ -385,7 +385,7 @@ impl Checksum {
        Checksum::default()
    }

-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn new_checksum_string(t: ChecksumMode, s: &str) -> Result<Checksum, std::io::Error> {
        let b = match base64_decode(s.as_bytes()) {
            Ok(b) => b,
@@ -412,7 +412,7 @@ impl Checksum {
        base64_encode(&self.r)
    }

-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn raw(&self) -> Option<Vec<u8>> {
        if !self.is_set() {
            return None;
@@ -37,16 +37,17 @@ pub struct PutObjReader {
    //pub sealMD5Fn: SealMD5CurrFn,
 }

-#[allow(dead_code)]
 impl PutObjReader {
    pub fn new(reader: HashReader) -> Self {
        Self { reader }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn md5_current_hex_string(&self) -> String {
        self.reader.checksum().map(|v| v.encoded).unwrap_or_default()
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn with_encryption(&mut self, enc_reader: HashReader) -> Result<(), std::io::Error> {
        self.reader = enc_reader;

@@ -54,6 +54,10 @@ use rustfs_config::MAX_S3_CLIENT_RESPONSE_SIZE;
 use rustfs_rio::HashReader;
 use rustfs_utils::HashAlgorithm;
 use rustfs_utils::{
+    http::headers::{
+        AMZ_CHECKSUM_CRC32, AMZ_CHECKSUM_CRC32C, AMZ_CHECKSUM_CRC64NVME, AMZ_CHECKSUM_MODE, AMZ_CHECKSUM_SHA1,
+        AMZ_CHECKSUM_SHA256,
+    },
    net::get_endpoint_url,
    retry::{DEFAULT_RETRY_CAP, DEFAULT_RETRY_UNIT, MAX_JITTER, MAX_RETRY, RetryTimer},
 };
@@ -1383,12 +1387,12 @@ pub(crate) fn to_object_info_for_provider(
    };

    // Extract checksums
-    let checksum_crc32 = get_header("x-amz-checksum-crc32");
-    let checksum_crc32c = get_header("x-amz-checksum-crc32c");
-    let checksum_sha1 = get_header("x-amz-checksum-sha1");
-    let checksum_sha256 = get_header("x-amz-checksum-sha256");
-    let checksum_crc64nvme = get_header("x-amz-checksum-crc64nvme");
-    let checksum_mode = get_header("x-amz-checksum-mode");
+    let checksum_crc32 = get_header(AMZ_CHECKSUM_CRC32);
+    let checksum_crc32c = get_header(AMZ_CHECKSUM_CRC32C);
+    let checksum_sha1 = get_header(AMZ_CHECKSUM_SHA1);
+    let checksum_sha256 = get_header(AMZ_CHECKSUM_SHA256);
+    let checksum_crc64nvme = get_header(AMZ_CHECKSUM_CRC64NVME);
+    let checksum_mode = get_header(AMZ_CHECKSUM_MODE);

    // Build and return the ObjectInfo struct
    Ok(ObjectInfo {
@@ -233,11 +233,17 @@ pub struct NsScannerCapabilityRequest {
 #[async_trait]
 pub trait InternodeDataTransport: Send + Sync + std::fmt::Debug {
    async fn open_read(&self, request: ReadStreamRequest) -> Result<FileReader>;
+    async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result<FileReader> {
+        self.open_read(request).await
+    }
    /// Opens an owned-chunk stream when this transport can retain receive-buffer
    /// ownership. `None` preserves the established `open_read` fallback.
    async fn open_read_chunks(&self, _request: ReadStreamRequest) -> Result<Option<ChunkReaderBox>> {
        Ok(None)
    }
+    async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result<Option<ChunkReaderBox>> {
+        self.open_read_chunks(request).await
+    }
    async fn open_write(&self, request: WriteStreamRequest) -> Result<FileWriter>;
    async fn open_walk_dir(&self, request: WalkDirStreamRequest) -> Result<FileReader>;
    async fn open_ns_scanner(&self, _request: NsScannerStreamRequest) -> Result<FileReader> {
@@ -269,6 +275,15 @@ impl InternodeDataTransport for TcpHttpInternodeDataTransport {
        ))
    }

+    async fn open_read_fresh(&self, request: ReadStreamRequest) -> Result<FileReader> {
+        let url = build_read_file_stream_url(&request);
+        let mut headers = json_headers();
+        build_auth_headers(&url, &Method::GET, &mut headers)?;
+        Ok(Box::new(
+            HttpReader::new_fresh_connection_with_stall_timeout(url, Method::GET, headers, None, request.stall_timeout).await?,
+        ))
+    }
+
    async fn open_read_chunks(&self, request: ReadStreamRequest) -> Result<Option<ChunkReaderBox>> {
        let url = build_read_file_stream_url(&request);
        let mut headers = json_headers();
@@ -278,6 +293,16 @@ impl InternodeDataTransport for TcpHttpInternodeDataTransport {
        )))
    }

+    async fn open_read_chunks_fresh(&self, request: ReadStreamRequest) -> Result<Option<ChunkReaderBox>> {
+        let url = build_read_file_stream_url(&request);
+        let mut headers = json_headers();
+        build_auth_headers(&url, &Method::GET, &mut headers)?;
+        Ok(Some(Box::new(
+            HttpChunkReader::new_fresh_connection_with_stall_timeout(url, Method::GET, headers, None, request.stall_timeout)
+                .await?,
+        )))
+    }
+
    async fn open_write(&self, request: WriteStreamRequest) -> Result<FileWriter> {
        let server_epoch = self.put_file_auth_capability(&request.endpoint).await?;
        let nonce = server_epoch.map(|_| Uuid::new_v4());
@@ -86,6 +86,25 @@ const PEER_REST_RECOVERY_MAX_BACKOFF: Duration = Duration::from_secs(30);
 const SCANNER_ACTIVITY_MAX_MESSAGE_SIZE: usize = 1024;
 const REPLICATION_STATS_MAX_MESSAGE_SIZE: usize = 8 * 1024 * 1024;

+/// Error for a peer that reported `success = false` without an `error_info` payload.
+///
+/// Same shape as `peer_s3_client::peer_failure_without_details`, over `StorageError`
+/// instead of `DiskError`. The message names the operation (and the bucket, where the
+/// operation has one) and nothing else, for two reasons:
+///
+/// - `finalize_result` classifies failures by message substring, so any text matching
+///   `message_has_network_needle` would take an answering peer offline and evict its
+///   connection over a plain application-level rejection.
+/// - Quorum aggregation (`reduce_errs`) buckets `Io` errors by kind plus rendered
+///   message, so a per-peer detail such as the peer address would split one shared
+///   failure into single-count buckets and downgrade the dominant error.
+fn peer_failure_without_details(op: &str, bucket: Option<&str>) -> Error {
+    match bucket {
+        Some(bucket) => Error::other(format!("{op}({bucket}): peer returned failure without error details")),
+        None => Error::other(format!("{op}: peer returned failure without error details")),
+    }
+}
+
 fn decode_bucket_stats_response(response: GetBucketStatsDataResponse) -> Result<BucketStats> {
    if !response.success {
        return Err(Error::other(
@@ -696,7 +715,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("local_storage_info", None));
        }
        let data = response.storage_info;

@@ -719,7 +738,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("server_info", None));
        }
        let data = response.server_properties;

@@ -742,7 +761,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("get_cpus", None));
        }
        let data = response.cpus;

@@ -765,7 +784,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("get_net_info", None));
        }
        let data = response.net_info;

@@ -788,7 +807,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("get_partitions", None));
        }
        let data = response.partitions;

@@ -811,7 +830,7 @@ impl PeerRestClient {
            if let Some(msg) = response.error_info {
                return Err(Error::other(msg));
            }
-            return Err(Error::other(""));
+            return Err(peer_failure_without_details("get_os_info", None));
        }
        let data = response.os_info;

@@ -832,7 +851,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_se_linux_info", None));
                }
                let data = response.sys_services;

@@ -857,7 +876,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_sys_config", None));
                }
                let data = response.sys_config;

@@ -882,7 +901,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_sys_errors", None));
                }
                let data = response.sys_errors;

@@ -907,7 +926,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_mem_info", None));
                }
                let data = response.mem_info;

@@ -939,7 +958,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_metrics", None));
                }
                let data = response.realtime_metrics;

@@ -964,7 +983,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_live_events", None));
                }

                Ok(PeerLiveEventsBatch {
@@ -989,7 +1008,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("get_proc_info", None));
                }
                let data = response.proc_info;

@@ -1016,7 +1035,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("start_profiling", None));
                }
                Ok(())
            }
@@ -1323,7 +1342,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_bucket_metadata", Some(bucket)));
                }
                Ok(())
            }
@@ -1346,7 +1365,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("delete_bucket_metadata", Some(bucket)));
                }
                Ok(())
            }
@@ -1369,7 +1388,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("delete_policy", None));
                }
                Ok(())
            }
@@ -1392,7 +1411,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_policy", None));
                }
                Ok(())
            }
@@ -1417,7 +1436,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_policy_mapping", None));
                }
                Ok(())
            }
@@ -1440,7 +1459,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("delete_user", None));
                }
                Ok(())
            }
@@ -1463,7 +1482,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("delete_service_account", None));
                }
                Ok(())
            }
@@ -1487,7 +1506,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_user", None));
                }
                Ok(())
            }
@@ -1510,7 +1529,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_service_account", None));
                }
                Ok(())
            }
@@ -1533,7 +1552,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_group", None));
                }
                Ok(())
            }
@@ -1554,7 +1573,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("reload_site_replication_config", None));
                }
                Ok(())
            }
@@ -1597,7 +1616,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("signal_service", None));
                }
                validate_signal_service_protocol(sig, sub_sys, response.protocol_version)?;
                Ok(response)
@@ -1667,7 +1686,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("reload_pool_meta", None));
                }

                Ok(())
@@ -1691,7 +1710,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("stop_rebalance", None));
                }

                Ok(())
@@ -1725,7 +1744,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("load_rebalance_meta", None));
                }

                Ok(())
@@ -1753,7 +1772,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("start_decommission", None));
                }

                Ok(())
@@ -1777,7 +1796,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("decommission_cancel", None));
                }

                Ok(())
@@ -1801,7 +1820,7 @@ impl PeerRestClient {
                    if let Some(msg) = response.error_info {
                        return Err(Error::other(msg));
                    }
-                    return Err(Error::other(""));
+                    return Err(peer_failure_without_details("clear_decommission", None));
                }

                Ok(())
@@ -1947,6 +1966,8 @@ fn tier_config_reload_status_outcome(status: tonic::Status) -> TierConfigReloadO
 mod tests {
    use super::*;
    use crate::config::com::STORAGE_CLASS_SUB_SYS;
+    use crate::disk::error::DiskError;
+    use crate::disk::error_reduce::reduce_errs;
    use crate::layout::{disks_layout::DisksLayout, endpoints::SetupType};
    use rustfs_config::{ENV_KUBERNETES_SERVICE_HOST, ENV_LOCAL_ENDPOINT_HOST, ENV_STARTUP_TOPOLOGY_WAIT_MODE};
    use serde_json::Value;
@@ -3098,4 +3119,115 @@ mod tests {
                && span.get("request_id").and_then(Value::as_str) == Some("req-peer-rest")
        }));
    }
+
+    /// Every operation name passed to `peer_failure_without_details` in this file.
+    const PEER_FAILURE_OPS: &[&str] = &[
+        "local_storage_info",
+        "server_info",
+        "get_cpus",
+        "get_net_info",
+        "get_partitions",
+        "get_os_info",
+        "get_se_linux_info",
+        "get_sys_config",
+        "get_sys_errors",
+        "get_mem_info",
+        "get_metrics",
+        "get_live_events",
+        "get_proc_info",
+        "start_profiling",
+        "load_bucket_metadata",
+        "delete_bucket_metadata",
+        "delete_policy",
+        "load_policy",
+        "load_policy_mapping",
+        "delete_user",
+        "delete_service_account",
+        "load_user",
+        "load_service_account",
+        "load_group",
+        "reload_site_replication_config",
+        "signal_service",
+        "reload_pool_meta",
+        "stop_rebalance",
+        "load_rebalance_meta",
+        "start_decommission",
+        "decommission_cancel",
+        "clear_decommission",
+    ];
+
+    #[test]
+    fn peer_failure_without_details_names_operation_and_bucket() {
+        for op in PEER_FAILURE_OPS {
+            let message = peer_failure_without_details(op, None).to_string();
+            assert!(message.contains(op), "{op} message must name the operation: {message}");
+        }
+
+        for op in ["load_bucket_metadata", "delete_bucket_metadata"] {
+            let message = peer_failure_without_details(op, Some("ops-bucket")).to_string();
+            assert!(message.contains(op), "{op} message must name the operation: {message}");
+            assert!(message.contains("ops-bucket"), "{op} message must name the bucket: {message}");
+        }
+    }
+
+    #[test]
+    fn peer_failure_without_details_keeps_one_reduce_errs_bucket_per_operation() {
+        // reduce_errs groups Io errors by kind plus rendered message: peers failing the
+        // same operation must stay a single dominant error instead of one bucket per peer.
+        let per_peer_errs = (0..4)
+            .map(|_| Some(DiskError::from(peer_failure_without_details("load_bucket_metadata", Some("shared")))))
+            .collect::<Vec<_>>();
+        let (count, dominant) = reduce_errs(&per_peer_errs, &[]);
+        assert_eq!(count, 4, "one shared failure must not split into per-peer buckets");
+        assert_eq!(
+            dominant,
+            Some(DiskError::from(peer_failure_without_details("load_bucket_metadata", Some("shared"))))
+        );
+
+        assert_ne!(
+            peer_failure_without_details("load_bucket_metadata", Some("shared")).to_string(),
+            peer_failure_without_details("delete_bucket_metadata", Some("shared")).to_string()
+        );
+        assert_ne!(
+            peer_failure_without_details("load_bucket_metadata", Some("bucket-a")).to_string(),
+            peer_failure_without_details("load_bucket_metadata", Some("bucket-b")).to_string()
+        );
+    }
+
+    #[test]
+    fn peer_failure_without_details_never_reads_as_a_network_failure() {
+        // `finalize_result` marks the peer offline and evicts its connection whenever the
+        // message matches a network needle. A peer that answered `success = false` is alive,
+        // so no operation or bucket name may push this text over that classifier.
+        for op in PEER_FAILURE_OPS {
+            let err = peer_failure_without_details(op, None);
+            assert!(
+                !PeerRestClient::is_network_like_error(&err),
+                "{op} must not read as a transport failure: {err}"
+            );
+
+            let scoped = peer_failure_without_details(op, Some("bucket-name"));
+            assert!(
+                !PeerRestClient::is_network_like_error(&scoped),
+                "{op} must not read as a transport failure: {scoped}"
+            );
+        }
+
+        // The bucket name is caller-supplied. Every needle carries a space, which S3 bucket
+        // names cannot, and the name is closed by `)` before the literal text resumes, so no
+        // needle can straddle the boundary either.
+        for bucket in [
+            "timed-out",
+            "connection-reset",
+            "transport-error",
+            "broken-pipe",
+            "unavailable-logs",
+        ] {
+            let err = peer_failure_without_details("load_bucket_metadata", Some(bucket));
+            assert!(
+                !PeerRestClient::is_network_like_error(&err),
+                "bucket {bucket} must not push the message over the network classifier: {err}"
+            );
+        }
+    }
 }
@@ -214,6 +214,21 @@ fn pool_write_quorum(participant_count: usize) -> usize {
    (participant_count / 2) + 1
 }

+/// Error for a peer that reported `success = false` without an error payload.
+///
+/// The message must stay identical across the peers of one operation: `reduce_errs`
+/// buckets `Error::Io` by kind plus rendered message, so any per-peer detail (address,
+/// timing) would split one shared failure into single-count buckets and downgrade a real
+/// dominant error into `ErasureWriteQuorum`.
+///
+/// `peer_rest_client` carries the same helper over `StorageError` for the same response shape.
+fn peer_failure_without_details(op: &str, bucket: Option<&str>) -> Error {
+    match bucket {
+        Some(bucket) => Error::other(format!("{op}({bucket}): peer returned failure without error details")),
+        None => Error::other(format!("{op}: peer returned failure without error details")),
+    }
+}
+
 fn reduce_pool_write_quorum_errs(per_pool_errs: &[Option<Error>]) -> Option<Error> {
    if per_pool_errs.is_empty() {
        return Some(Error::ErasureWriteQuorum);
@@ -1078,7 +1093,7 @@ impl PeerS3Client for RemotePeerS3Client {
                    return if let Some(err) = response.error {
                        Err(err.into())
                    } else {
-                        Err(Error::other(""))
+                        Err(peer_failure_without_details("heal_bucket", Some(bucket)))
                    };
                }

@@ -1105,7 +1120,7 @@ impl PeerS3Client for RemotePeerS3Client {
                    return if let Some(err) = response.error {
                        Err(err.into())
                    } else {
-                        Err(Error::other(""))
+                        Err(peer_failure_without_details("list_bucket", None))
                    };
                }
                let bucket_infos = response
@@ -1136,9 +1151,7 @@ impl PeerS3Client for RemotePeerS3Client {
                    return if let Some(err) = response.error {
                        Err(err.into())
                    } else {
-                        Err(Error::other(format!(
-                            "make_bucket({bucket}): peer returned failure without error details"
-                        )))
+                        Err(peer_failure_without_details("make_bucket", Some(bucket)))
                    };
                }

@@ -1162,7 +1175,7 @@ impl PeerS3Client for RemotePeerS3Client {
                    return if let Some(err) = response.error {
                        Err(err.into())
                    } else {
-                        Err(Error::other(""))
+                        Err(peer_failure_without_details("get_bucket_info", Some(bucket)))
                    };
                }
                let bucket_info = serde_json::from_str::<BucketInfo>(&response.bucket_info)?;
@@ -1190,7 +1203,7 @@ impl PeerS3Client for RemotePeerS3Client {
                    return if let Some(err) = response.error {
                        Err(err.into())
                    } else {
-                        Err(Error::other(""))
+                        Err(peer_failure_without_details("delete_bucket", Some(bucket)))
                    };
                }

@@ -2314,4 +2327,37 @@ mod tests {
            .collect::<Vec<_>>();
        assert_eq!(calls, vec![1, 1, 0, 0, 0, 0, 0, 0]);
    }
+
+    #[test]
+    fn peer_failure_without_details_names_operation_and_bucket() {
+        for op in ["heal_bucket", "make_bucket", "get_bucket_info", "delete_bucket"] {
+            let message = peer_failure_without_details(op, Some("ops-bucket")).to_string();
+            assert!(message.contains(op), "{op} message must name the operation: {message}");
+            assert!(message.contains("ops-bucket"), "{op} message must name the bucket: {message}");
+        }
+
+        let message = peer_failure_without_details("list_bucket", None).to_string();
+        assert!(message.contains("list_bucket"), "cluster-wide message must name the operation");
+        assert!(!message.trim().is_empty());
+    }
+
+    #[test]
+    fn peer_failure_without_details_keeps_one_reduce_errs_bucket_per_operation() {
+        // reduce_errs groups Io errors by kind plus rendered message: peers failing the
+        // same operation on the same bucket must still reach quorum as one dominant error.
+        let per_pool_errs = vec![
+            Some(peer_failure_without_details("delete_bucket", Some("shared"))),
+            Some(peer_failure_without_details("delete_bucket", Some("shared"))),
+            Some(peer_failure_without_details("delete_bucket", Some("shared"))),
+        ];
+        assert_eq!(
+            reduce_pool_write_quorum_errs(&per_pool_errs),
+            Some(peer_failure_without_details("delete_bucket", Some("shared")))
+        );
+
+        assert_ne!(
+            peer_failure_without_details("delete_bucket", Some("shared")),
+            peer_failure_without_details("get_bucket_info", Some("shared"))
+        );
+    }
 }
@@ -39,7 +39,6 @@ use rustfs_config::{
 };
 use std::sync::LazyLock;

-#[allow(dead_code)]
 #[allow(clippy::declare_interior_mutable_const)]
 /// Default KVS for audit webhook settings.
 pub static DEFAULT_AUDIT_WEBHOOK_KVS: LazyLock<KVS> = LazyLock::new(|| {
@@ -117,7 +116,6 @@ pub static DEFAULT_AUDIT_WEBHOOK_KVS: LazyLock<KVS> = LazyLock::new(|| {
    ])
 });

-#[allow(dead_code)]
 #[allow(clippy::declare_interior_mutable_const)]
 /// Default KVS for audit MQTT settings.
 pub static DEFAULT_AUDIT_MQTT_KVS: LazyLock<KVS> = LazyLock::new(|| {
@@ -375,7 +373,6 @@ pub static DEFAULT_AUDIT_NATS_KVS: LazyLock<KVS> = LazyLock::new(|| {
    ])
 });

-#[allow(dead_code)]
 pub static DEFAULT_AUDIT_PULSAR_KVS: LazyLock<KVS> = LazyLock::new(|| {
    KVS(vec![
        KV {
@@ -12,12 +12,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.

-use crate::error::{Error, Result};
 use rustfs_config::server_config::{KV, KVS};
 use rustfs_config::{DEFAULT_HEAL_BITROT_CYCLE_SECS, HEAL_BITROT_CYCLE};
-use rustfs_utils::string::parse_bool;
 use std::sync::LazyLock;
-use std::time::Duration;

 pub static DEFAULT_KVS: LazyLock<KVS> = LazyLock::new(|| {
    KVS(vec![KV {
@@ -26,59 +23,3 @@ pub static DEFAULT_KVS: LazyLock<KVS> = LazyLock::new(|| {
        hidden_if_empty: false,
    }])
 });
-
-#[derive(Debug, Default)]
-pub struct Config {
-    pub bitrot: String,
-    pub sleep: Duration,
-    pub io_count: usize,
-    pub drive_workers: usize,
-    pub cache: Duration,
-}
-
-impl Config {
-    pub fn bitrot_scan_cycle(&self) -> Duration {
-        self.cache
-    }
-
-    pub fn get_workers(&self) -> usize {
-        self.drive_workers
-    }
-
-    pub fn update(&mut self, nopts: &Config) {
-        self.bitrot = nopts.bitrot.clone();
-        self.io_count = nopts.io_count;
-        self.sleep = nopts.sleep;
-        self.drive_workers = nopts.drive_workers;
-    }
-}
-
-const RUSTFS_BITROT_CYCLE_IN_MONTHS: u64 = 1;
-
-fn parse_bitrot_config(s: &str) -> Result<Duration> {
-    match parse_bool(s) {
-        Ok(enabled) => {
-            if enabled {
-                Ok(Duration::from_secs_f64(0.0))
-            } else {
-                Ok(Duration::from_secs_f64(-1.0))
-            }
-        }
-        Err(_) => {
-            if !s.ends_with("m") {
-                return Err(Error::other("unknown format"));
-            }
-
-            match s.trim_end_matches('m').parse::<u64>() {
-                Ok(months) => {
-                    if months < RUSTFS_BITROT_CYCLE_IN_MONTHS {
-                        return Err(Error::other(format!("minimum bitrot cycle is {RUSTFS_BITROT_CYCLE_IN_MONTHS} month(s)")));
-                    }
-
-                    Ok(Duration::from_secs(months * 30 * 24 * 60))
-                }
-                Err(err) => Err(Error::other(err)),
-            }
-        }
-    }
-}
@@ -16,7 +16,6 @@

 mod audit;
 pub mod com;
-#[allow(dead_code)]
 pub mod heal;
 mod notify;
 mod oidc;
@@ -16,6 +16,7 @@ use crate::bucket::replication::replication_state_from_filemeta;
 use crate::bucket::versioning_sys::BucketVersioningSys;
 use crate::bucket::{
    lifecycle::{
+        LifecycleExpiryConfigs,
        bucket_lifecycle_audit::LcEventSrc,
        bucket_lifecycle_ops::{
            LifecycleOps, apply_expiry_on_transitioned_object, apply_expiry_rule_in, eval_action_from_lifecycle,
@@ -1996,11 +1997,11 @@ impl PoolMeta {
        Ok(false)
    }

-    #[allow(dead_code)]
    pub fn validate(&self, pools: Vec<Arc<Sets>>) -> Result<bool> {
        struct PoolInfo {
            position: usize,
            completed: bool,
+            #[allow(dead_code, reason = "written but never read back (backlog#1823)")]
            decom_started: bool,
        }

@@ -2335,6 +2336,10 @@ fn lifecycle_action_removes_data_movement_version(action: IlmAction) -> bool {
    )
 }

+fn lifecycle_action_skips_heal_version(action: IlmAction) -> bool {
+    action.delete()
+}
+
 fn resolve_data_movement_lifecycle_expiry_result(action: IlmAction, apply_actions: bool, applied: bool) -> Result<bool> {
    if !apply_actions || applied {
        return Ok(true);
@@ -2385,7 +2390,80 @@ pub(crate) async fn should_skip_lifecycle_for_data_movement(
    }
 }

+pub struct HealLifecycleExpiryContext {
+    configs: LifecycleExpiryConfigs,
+}
+
 impl ECStore {
+    pub async fn load_heal_lifecycle_expiry_context(&self, bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+        if bucket == RUSTFS_META_BUCKET {
+            return Ok(None);
+        }
+
+        let configs = get_expiry_configs(self, bucket).await?;
+        if configs.lifecycle.is_none() {
+            return Ok(None);
+        }
+
+        Ok(Some(HealLifecycleExpiryContext { configs }))
+    }
+
+    pub async fn enqueue_heal_lifecycle_expiry(
+        self: &Arc<Self>,
+        context: &HealLifecycleExpiryContext,
+        bucket: &str,
+        object: &str,
+        version_id: Option<&str>,
+        object_info: Option<&crate::object_api::ObjectInfo>,
+    ) -> Result<bool> {
+        let Some(lifecycle_config) = context.configs.lifecycle.as_ref() else {
+            return Ok(false);
+        };
+
+        let object_info = if let Some(object_info) = object_info {
+            if object_info.bucket != bucket || object_info.name != object {
+                return Ok(false);
+            }
+            let snapshot_version_id = object_info
+                .version_id
+                .filter(|version_id| !version_id.is_nil())
+                .map(|version_id| version_id.to_string());
+            if snapshot_version_id.as_deref() != version_id {
+                return Ok(false);
+            }
+            object_info.clone()
+        } else {
+            match self
+                .get_object_info(
+                    bucket,
+                    object,
+                    &ObjectOptions {
+                        version_id: version_id.map(str::to_string),
+                        versioned: version_id.is_some(),
+                        expected_bucket_incarnation_id: Some(context.configs.bucket_incarnation_id),
+                        ..Default::default()
+                    },
+                )
+                .await
+            {
+                Ok(object_info) => object_info,
+                Err(err) if is_err_object_not_found(&err) || is_err_version_not_found(&err) => return Ok(false),
+                Err(err) => return Err(err),
+            }
+        };
+
+        let event = eval_action_from_lifecycle(lifecycle_config, context.configs.object_lock.as_deref(), &object_info).await;
+        if !lifecycle_action_skips_heal_version(event.action) {
+            return Ok(false);
+        }
+
+        if lifecycle_delete_all_versions_blocked_by_replication(self.clone(), bucket, &object_info.name, event.action).await? {
+            return Ok(false);
+        }
+
+        Ok(apply_expiry_rule_in(self.clone(), &event, &LcEventSrc::Scanner, &object_info).await)
+    }
+
    async fn save_current_pool_meta(&self) -> Result<()> {
        let _save_guard = self.pool_meta_save_gate.lock().await;
        let snapshot = {
@@ -4287,6 +4365,19 @@ mod tests {
        ));
    }

+    #[test]
+    fn lifecycle_action_skips_heal_version_for_every_delete_action() {
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteVersionAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteRestoredAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteRestoredVersionAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DeleteAllVersionsAction));
+        assert!(lifecycle_action_skips_heal_version(IlmAction::DelMarkerDeleteAllVersionsAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::TransitionAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::TransitionVersionAction));
+        assert!(!lifecycle_action_skips_heal_version(IlmAction::NoneAction));
+    }
+
    #[test]
    fn resolve_data_movement_lifecycle_expiry_result_allows_dry_run_skip() {
        let skip = resolve_data_movement_lifecycle_expiry_result(IlmAction::DeleteVersionAction, false, false)
@@ -4958,13 +5049,19 @@ fn is_disk_online_state(state: &str) -> bool {
 }

 #[deprecated(since = "0.1.0", note = "Use fallback_total_capacity_dedup instead")]
-#[allow(dead_code)]
+#[allow(
+    dead_code,
+    reason = "superseded by the replacement named in the comment at pools.rs:5071 (backlog#1823)"
+)]
 fn fallback_total_capacity(disks: &[rustfs_madmin::Disk]) -> usize {
    fallback_total_capacity_dedup(disks)
 }

 #[deprecated(since = "0.1.0", note = "Use fallback_free_capacity_dedup instead")]
-#[allow(dead_code)]
+#[allow(
+    dead_code,
+    reason = "superseded by the replacement named in the comment at pools.rs:5071 (backlog#1823)"
+)]
 fn fallback_free_capacity(disks: &[rustfs_madmin::Disk]) -> usize {
    fallback_free_capacity_dedup(disks)
 }
@@ -1140,11 +1140,11 @@ impl crate::storage_api_contracts::heal::HealOperations for Sets {

        Err(Error::DiskNotFound)
    }
-    #[tracing::instrument(skip(self))]
-    async fn check_abandoned_parts(&self, _bucket: &str, _object: &str, _opts: &HealOpts) -> Result<()> {
-        // Multipart orphan reconciliation is intentionally retained above the pool/set layers
-        // until there is a concrete caller and a stable lower-level contract to implement.
-        Err(StorageError::NotImplemented)
+    #[tracing::instrument(level = "debug", skip(self, opts), fields(bucket = %bucket, object = %object, dry_run = opts.dry_run))]
+    async fn check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
+        self.get_disks_for_heal_object(object, opts)?
+            .check_abandoned_parts(bucket, object, opts)
+            .await
    }
 }

@@ -1996,7 +1996,7 @@ mod tests {
    }

    #[tokio::test]
-    async fn sets_check_abandoned_parts_returns_typed_not_implemented_error() {
+    async fn sets_check_abandoned_parts_rejects_invalid_set_scope() {
        let format = FormatV3::new(1, 1);
        let sets = Sets {
            id: format.id,
@@ -2021,10 +2021,21 @@ mod tests {
        };

        let err = sets
-            .check_abandoned_parts("bucket", "object", &HealOpts::default())
+            .check_abandoned_parts(
+                "bucket",
+                "object",
+                &HealOpts {
+                    set: Some(1),
+                    ..Default::default()
+                },
+            )
            .await
-            .expect_err("abandoned-parts ownership should stay above the pool/set storage layers");
-        assert!(matches!(err, StorageError::NotImplemented));
+            .expect_err("out-of-range abandoned-parts set scope must fail closed");
+        assert!(
+            matches!(err, StorageError::InvalidArgument(_, ref field, ref reason)
+                if field == "set" && reason.contains("invalid heal set index 1")),
+            "unexpected invalid set error: {err:?}"
+        );
    }

    // Builds a single-set `Sets` over `SET_DRIVE_COUNT` local temp-dir disks,
@@ -418,6 +418,17 @@ pub struct DiskHealthTracker {
    pub last_capacity_free: AtomicU64,
    /// Last successful capacity probe timestamp
    pub last_capacity_probe_unix_secs: AtomicI64,
+    /// Authoritative atomically published runtime/status pair.
+    state_snapshot: AtomicU64,
+    transition_lock: std::sync::Mutex<()>,
+}
+
+fn pack_health_state(runtime_state: RuntimeDriveHealthState, status: u32) -> u64 {
+    (u64::from(runtime_state as u32) << 32) | u64::from(status)
+}
+
+fn unpack_health_state(snapshot: u64) -> (RuntimeDriveHealthState, u32) {
+    (RuntimeDriveHealthState::from_u32((snapshot >> 32) as u32), snapshot as u32)
 }

 #[derive(Debug)]
@@ -637,14 +648,23 @@ impl Default for DiskOperationMetrics {
 }

 impl DiskOperationMetrics {
+    #[allow(
+        dead_code,
+        reason = "internal metrics recorder reached only from record() below (backlog#1823)"
+    )]
    fn record_call(&mut self) {
        self.lifetime_calls.fetch_add(1, Ordering::Relaxed);
    }

+    #[allow(
+        dead_code,
+        reason = "internal metrics recorder reached only from record() below (backlog#1823)"
+    )]
    fn record_latency(&mut self, now_sec: u64, elapsed: Duration) {
        self.record_latency_atomic(now_sec, elapsed);
    }

+    #[allow(dead_code, reason = "metrics roll-up with no caller in this port (backlog#1823)")]
    fn record(&mut self, now_sec: u64, elapsed: Duration) {
        self.record_call();
        self.record_latency(now_sec, elapsed);
@@ -730,6 +750,8 @@ impl DiskHealthTracker {
            last_capacity_used: AtomicU64::new(0),
            last_capacity_free: AtomicU64::new(0),
            last_capacity_probe_unix_secs: AtomicI64::new(0),
+            state_snapshot: AtomicU64::new(pack_health_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK)),
+            transition_lock: std::sync::Mutex::new(()),
        }
    }

@@ -766,38 +788,52 @@ impl DiskHealthTracker {

    /// Check if disk is faulty
    pub fn is_faulty(&self) -> bool {
-        self.status.load(Ordering::Acquire) == DISK_HEALTH_FAULTY
+        unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1 == DISK_HEALTH_FAULTY
+    }
+
+    fn publish_state(&self, runtime_state: RuntimeDriveHealthState, status: u32) {
+        self.state_snapshot
+            .store(pack_health_state(runtime_state, status), Ordering::Release);
+        self.runtime_state.store(runtime_state as u32, Ordering::Release);
+        self.status.store(status, Ordering::Release);
    }

    /// Set disk as faulty
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub fn set_faulty(&self) {
-        self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY);
    }

    /// Set disk as OK
    pub fn set_ok(&self) {
-        self.status.store(DISK_HEALTH_OK, Ordering::Release);
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK);
    }

    #[cfg(test)]
    pub fn force_runtime_state_for_test(&self, state: RuntimeDriveHealthState) {
-        self.runtime_state.store(state as u32, Ordering::Release);
-        match state {
-            RuntimeDriveHealthState::Offline => self.set_faulty(),
-            RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect | RuntimeDriveHealthState::Returning => {
-                self.set_ok();
-            }
-        }
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        let status = if state == RuntimeDriveHealthState::Offline {
+            DISK_HEALTH_FAULTY
+        } else {
+            DISK_HEALTH_OK
+        };
+        self.publish_state(state, status);
    }

    pub fn swap_ok_to_faulty(&self) -> bool {
-        self.status
-            .compare_exchange(DISK_HEALTH_OK, DISK_HEALTH_FAULTY, Ordering::AcqRel, Ordering::Relaxed)
-            .is_ok()
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
+        let (_, status) = unpack_health_state(self.state_snapshot.load(Ordering::Acquire));
+        if status != DISK_HEALTH_OK {
+            return false;
+        }
+        self.publish_state(RuntimeDriveHealthState::Offline, DISK_HEALTH_FAULTY);
+        true
    }

    pub fn runtime_state(&self) -> RuntimeDriveHealthState {
-        RuntimeDriveHealthState::from_u32(self.runtime_state.load(Ordering::Acquire))
+        unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).0
    }

    pub fn offline_duration(&self) -> Option<Duration> {
@@ -813,6 +849,7 @@ impl DiskHealthTracker {
    }

    pub fn mark_failure(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
        let current = self.runtime_state();
        let now = current_unix_secs();
        let next = match current {
@@ -841,23 +878,19 @@ impl DiskHealthTracker {
        };

        let became_offline = next == RuntimeDriveHealthState::Offline && current != RuntimeDriveHealthState::Offline;
-        if next == RuntimeDriveHealthState::Offline {
-            self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
-        } else {
-            self.status.store(DISK_HEALTH_OK, Ordering::Release);
-        }
        self.transition_state(endpoint, current, next, reason);
        became_offline
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub fn mark_offline(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
        let current = self.runtime_state();
        if current == RuntimeDriveHealthState::Offline {
            return false;
        }

        self.consecutive_successes.store(0, Ordering::Release);
-        self.status.store(DISK_HEALTH_FAULTY, Ordering::Release);
        self.transition_state(endpoint, current, RuntimeDriveHealthState::Offline, reason);
        true
    }
@@ -871,11 +904,10 @@ impl DiskHealthTracker {
    }

    fn reset_for_store_init_retry_at(&self, endpoint: &Endpoint, now: Duration) {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
        let now_nanos = unix_nanos(now);
        let now_secs = unix_secs_i64(now);
-        self.status.store(DISK_HEALTH_OK, Ordering::Release);
-        self.runtime_state
-            .store(RuntimeDriveHealthState::Online as u32, Ordering::Release);
+        self.publish_state(RuntimeDriveHealthState::Online, DISK_HEALTH_OK);
        self.consecutive_failures.store(0, Ordering::Release);
        self.consecutive_successes.store(0, Ordering::Release);
        self.offline_since_unix_secs.store(0, Ordering::Release);
@@ -887,6 +919,7 @@ impl DiskHealthTracker {
    }

    pub fn mark_recovery_success(&self, endpoint: &Endpoint, reason: &'static str) -> bool {
+        let _guard = self.transition_lock.lock().unwrap_or_else(|poisoned| poisoned.into_inner());
        let current = self.runtime_state();
        let next = match current {
            RuntimeDriveHealthState::Online => RuntimeDriveHealthState::Online,
@@ -907,7 +940,6 @@ impl DiskHealthTracker {

        let became_online = next == RuntimeDriveHealthState::Online;
        if became_online {
-            self.status.store(DISK_HEALTH_OK, Ordering::Release);
            self.consecutive_failures.store(0, Ordering::Release);
            self.consecutive_successes.store(0, Ordering::Release);
        }
@@ -937,7 +969,13 @@ impl DiskHealthTracker {
            return;
        }

-        self.runtime_state.store(next as u32, Ordering::Release);
+        let current_status = unpack_health_state(self.state_snapshot.load(Ordering::Acquire)).1;
+        let status = match next {
+            RuntimeDriveHealthState::Offline => DISK_HEALTH_FAULTY,
+            RuntimeDriveHealthState::Returning => current_status,
+            RuntimeDriveHealthState::Online | RuntimeDriveHealthState::Suspect => DISK_HEALTH_OK,
+        };
+        self.publish_state(next, status);
        self.last_transition_unix_secs
            .store(current_unix_secs() as i64, Ordering::Release);

@@ -980,11 +1018,13 @@ impl DiskHealthTracker {
    }

    /// Get waiting operations count
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub fn waiting_count(&self) -> u32 {
        self.waiting.load(Ordering::Relaxed)
    }

    /// Get last success timestamp
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub fn last_success(&self) -> i64 {
        self.last_success.load(Ordering::Acquire)
    }
@@ -1026,21 +1066,6 @@ impl Default for DiskHealthTracker {
    }
 }

-/// Health check context key for tracking disk operations
-#[derive(Debug, Clone)]
-struct HealthDiskCtxKey;
-
-#[derive(Debug)]
-struct HealthDiskCtxValue {
-    last_success: Arc<AtomicI64>,
-}
-
-impl HealthDiskCtxValue {
-    fn log_success(&self) {
-        self.last_success.store(current_unix_nanos(), Ordering::Relaxed);
-    }
-}
-
 /// LocalDiskWrapper wraps a DiskStore with health tracking capabilities.
 /// This is similar to Go's xlStorageDiskIDCheck.
 #[derive(Debug, Clone)]
@@ -1072,10 +1097,6 @@ impl LocalDiskWrapper {
        )
    }

-    pub(crate) fn new_with_health(disk: Arc<LocalDisk>, health_check: bool, health: Arc<DiskHealthTracker>) -> Self {
-        Self::new_with_health_and_metrics(disk, health_check, health, Arc::new(DiskHealthMetricEpoch::default()))
-    }
-
    pub(crate) fn new_with_reconnect_state(
        disk: Arc<LocalDisk>,
        health_check: bool,
@@ -1223,7 +1244,7 @@ impl LocalDiskWrapper {
                        return;
                    }

-                    if health.status.load(Ordering::Relaxed) != DISK_HEALTH_OK {
+                    if health.is_faulty() {
                        continue;
                    }

@@ -1438,20 +1459,6 @@ impl LocalDiskWrapper {
        }
    }

-    async fn check_id(&self, want_id: Option<Uuid>) -> Result<()> {
-        if want_id.is_none() {
-            return Ok(());
-        }
-
-        let stored_disk_id = self.disk.get_disk_id().await?;
-
-        if stored_disk_id != want_id {
-            return Err(Error::other(format!("Disk ID mismatch wanted {want_id:?}, got {stored_disk_id:?}")));
-        }
-
-        Ok(())
-    }
-
    /// Check if disk ID is stale
    async fn check_disk_stale(&self) -> Result<()> {
        let Some(current_disk_id) = *self.disk_id.read().await else {
@@ -2929,6 +2936,57 @@ mod tests {
        });
    }

+    #[test]
+    #[serial_test::serial]
+    fn concurrent_failure_and_recovery_publish_one_health_snapshot() {
+        temp_env::with_var(rustfs_config::ENV_DRIVE_SUSPECT_FAILURE_THRESHOLD, Some("2"), || {
+            let endpoint = Endpoint::try_from("/tmp/concurrent-health-snapshot").expect("endpoint should parse");
+            let health = Arc::new(DiskHealthTracker::new());
+            let transition_guard = health
+                .transition_lock
+                .lock()
+                .expect("health transition lock should not be poisoned");
+            let start = Arc::new(std::sync::Barrier::new(3));
+            let (completed_tx, completed_rx) = std::sync::mpsc::channel();
+            let workers = (0..2)
+                .map(|_| {
+                    let health = Arc::clone(&health);
+                    let endpoint = endpoint.clone();
+                    let start = Arc::clone(&start);
+                    let completed_tx = completed_tx.clone();
+                    std::thread::spawn(move || {
+                        start.wait();
+                        health.mark_failure(&endpoint, "concurrent_test");
+                        completed_tx.send(()).expect("completion receiver should remain available");
+                    })
+                })
+                .collect::<Vec<_>>();
+
+            start.wait();
+            assert!(
+                matches!(
+                    completed_rx.recv_timeout(Duration::from_millis(250)),
+                    Err(std::sync::mpsc::RecvTimeoutError::Timeout)
+                ),
+                "concurrent transitions must wait for the serialization lock"
+            );
+            drop(transition_guard);
+            completed_rx
+                .recv_timeout(Duration::from_secs(1))
+                .expect("first failure transition should complete after lock release");
+            completed_rx
+                .recv_timeout(Duration::from_secs(1))
+                .expect("second failure transition should complete after lock release");
+            for worker in workers {
+                worker.join().expect("health transition worker should not panic");
+            }
+
+            assert_eq!(health.runtime_state(), RuntimeDriveHealthState::Offline);
+            assert!(health.is_faulty());
+            assert_eq!(health.consecutive_failures.load(Ordering::Acquire), 2);
+        });
+    }
+
    #[test]
    fn operation_success_recovers_suspect_drive_without_faulting() {
        let endpoint = Endpoint::try_from("/tmp/runtime-state-suspect-success").expect("endpoint should parse");
@@ -48,6 +48,7 @@ pub fn to_volume_error(io_err: std::io::Error) -> std::io::Error {
    }
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn to_disk_error(io_err: std::io::Error) -> std::io::Error {
    match io_err.kind() {
        std::io::ErrorKind::NotFound => DiskError::DiskNotFound.into(),
@@ -178,6 +178,7 @@ pub async fn remove(path: impl AsRef<Path>) -> io::Result<()> {
    }
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub async fn remove_all(path: impl AsRef<Path>) -> io::Result<()> {
    // Try remove_file first; fall back to remove_dir_all if it's a directory
    match fs::remove_file(path.as_ref()).await {
@@ -665,6 +665,7 @@ async fn remove_empty_directory_tree_under_mount_lease(
 }

 #[cfg(unix)]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 async fn remove_empty_directory_tree_with(
    root: &Path,
    before_descend: impl FnMut(&Path) -> std::io::Result<()>,
@@ -1016,13 +1017,29 @@ fn record_direct_read_page_fault_delta(path: &'static str, stage: &'static str,
 /// When enabled, shard reads bypass the page cache using O_DIRECT flag.
 /// Requires aligned buffers (typically 512 bytes or 4096 bytes).
 /// Default: false (uses page cache via mmap/pread).
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const ENV_RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE: &str = "RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE";
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const DEFAULT_RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE: bool = false;

 /// Minimum shard size threshold for O_DIRECT reads.
 /// Only shards larger than this threshold will use O_DIRECT.
 /// Default: 4MB.
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const ENV_RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD: &str = "RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD";
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const DEFAULT_RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD: usize = 4 * 1024 * 1024;

 /// Enable O_DIRECT for erasure shard / multipart part data writes (Linux only).
@@ -1036,7 +1053,15 @@ const DEFAULT_RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD: usize = 4 * 1024 * 1024;
 /// EINVAL/EOPNOTSUPP (tmpfs, overlayfs, 9p, ...) latch the path off and fall
 /// back to buffered writes for the whole disk. Non-Linux always falls back.
 /// Default: false (buffered writes via the page cache, as before).
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const ENV_RUSTFS_OBJECT_DIRECT_IO_WRITE_ENABLE: &str = "RUSTFS_OBJECT_DIRECT_IO_WRITE_ENABLE";
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const DEFAULT_RUSTFS_OBJECT_DIRECT_IO_WRITE_ENABLE: bool = false;
 const ENV_RUSTFS_OBJECT_MMAP_POPULATE_ENABLE: &str = "RUSTFS_OBJECT_MMAP_POPULATE_ENABLE";
 const DEFAULT_RUSTFS_OBJECT_MMAP_POPULATE_ENABLE: bool = false;
@@ -1095,12 +1120,14 @@ macro_rules! cached_read_env {

 cached_read_env! {
    /// Check if O_DIRECT reads are enabled.
+    #[allow(dead_code, reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)")]
    fn is_direct_io_read_enabled() -> bool =
        rustfs_utils::get_env_bool(ENV_RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE, DEFAULT_RUSTFS_OBJECT_DIRECT_IO_READ_ENABLE);
 }

 cached_read_env! {
    /// Check if O_DIRECT shard/part data writes are enabled.
+    #[allow(dead_code, reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)")]
    fn is_direct_io_write_enabled() -> bool =
        rustfs_utils::get_env_bool(ENV_RUSTFS_OBJECT_DIRECT_IO_WRITE_ENABLE, DEFAULT_RUSTFS_OBJECT_DIRECT_IO_WRITE_ENABLE);
 }
@@ -1456,6 +1483,7 @@ pub(crate) fn effective_durability(volume: &str) -> DurabilityMode {

 cached_read_env! {
    /// Get the O_DIRECT read threshold size.
+    #[allow(dead_code, reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)")]
    fn get_direct_io_read_threshold() -> usize =
        rustfs_utils::get_env_usize(ENV_RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD, DEFAULT_RUSTFS_OBJECT_DIRECT_IO_READ_THRESHOLD);
 }
@@ -1673,12 +1701,20 @@ impl DirectIoWriteState {
 /// Target staging size for O_DIRECT writes, rounded up to the DIO alignment.
 /// Bounds the per-writer aligned bounce buffer and batches many shard blocks
 /// into one positioned write to keep the syscall count low.
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 const DIRECT_WRITE_STAGING_BYTES: usize = 1024 * 1024;

 /// Aligned bounce-buffer capacity for a given DIO alignment: the target staging
 /// size rounded up to a whole multiple of `align` so the buffer address, every
 /// flushed batch length, and every write offset stay alignment-correct.
 /// Platform-independent (no O_DIRECT), so it is unit-tested on any host.
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 fn direct_write_staging_capacity(align: usize) -> usize {
    debug_assert!(align.is_power_of_two() && align >= 512);
    DIRECT_WRITE_STAGING_BYTES.div_ceil(align) * align
@@ -1687,6 +1723,10 @@ fn direct_write_staging_capacity(align: usize) -> usize {
 /// Split `filled` staged bytes into the alignment-sized prefix written with
 /// O_DIRECT and the sub-alignment tail written buffered. Platform-independent,
 /// so the tail-boundary math is unit-tested on any host.
+#[allow(
+    dead_code,
+    reason = "platform-conditional: production callers are inside #[cfg(target_os = \"linux\")] blocks, so this reads as dead on non-Linux hosts (backlog#1823)"
+)]
 fn direct_write_tail_split(filled: usize, align: usize) -> (usize, usize) {
    let aligned = filled - (filled % align);
    (aligned, filled - aligned)
@@ -2142,6 +2182,7 @@ fn set_delete_version_fail_after_data_staged(path: &str) {
 }

 #[cfg(test)]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(crate) fn set_delete_version_fail_after_commit(root: &Path, path: &str) {
    DELETE_VERSION_FAIL_AFTER_COMMIT
        .lock()
@@ -2447,6 +2488,10 @@ enum SyncMode {
    FileOnly,
 }

+#[allow(
+    dead_code,
+    reason = "reclaim bookkeeping fields written by Drop but never read back (backlog#1823)"
+)]
 struct FileCacheReclaimWriter {
    inner: File,
    reclaim_len: usize,
@@ -2454,6 +2499,10 @@ struct FileCacheReclaimWriter {
    reclaimed: bool,
 }

+#[allow(
+    dead_code,
+    reason = "reclaim bookkeeping fields written by Drop but never read back (backlog#1823)"
+)]
 struct FileCacheReclaimReader {
    inner: File,
    reclaim_offset: u64,
@@ -2519,6 +2568,10 @@ impl<R: AsyncRead + Unpin> AsyncRead for StallTimeoutReader<R> {
    }
 }

+#[allow(
+    dead_code,
+    reason = "reclaim metrics emitter reached only from the Linux-gated reclaim paths (backlog#1823)"
+)]
 fn record_file_cache_reclaim_success(kind: &'static str, reclaim_len: usize, started: std::time::Instant) {
    // Runs per read-stream page-cache reclaim window; skip the whole emission
    // (three metric-key constructions) when general metrics are disabled.
@@ -3071,6 +3124,7 @@ impl LocalIoBackend for StdBackend {
            use memmap2::MmapOptions;
            use std::time::{Duration as StdDuration, Instant as StdInstant};

+            #[allow(dead_code, reason = "mmap copy result slot kept beside the mapping it owns (backlog#1823)")]
            struct MmapCopyReadResult {
                bytes: Bytes,
                access_check_duration: StdDuration,
@@ -4704,6 +4758,10 @@ fn build_local_io_backend(root: PathBuf) -> Arc<dyn LocalIoBackend> {
    Arc::new(StdBackend::new(root))
 }

+#[allow(
+    dead_code,
+    reason = "path cache and cwd slots retained beside the disk root they derive from (backlog#1823)"
+)]
 pub struct LocalDisk {
    pub root: PathBuf,
    publication_root: os::PublicationRoot,
@@ -5490,6 +5548,7 @@ impl LocalDisk {
        Ok(Self::resolve_abs_path_from(&self.root, path.as_ref()))
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    fn io_resolve_abs_path(&self, path: impl AsRef<Path>) -> PathBuf {
        let path_ref = path.as_ref();
        let path_str = path_ref.to_string_lossy();
@@ -5567,15 +5626,24 @@ impl LocalDisk {
    }

    // Check if a path is valid
+    #[allow(
+        dead_code,
+        reason = "method wrapper over the live free function check_local_disk_valid_path; no caller in this port (backlog#1823)"
+    )]
    fn check_valid_path<P: AsRef<Path>>(&self, path: P) -> Result<()> {
        check_local_disk_valid_path(self.io_root(), path)
    }

+    #[allow(
+        dead_code,
+        reason = "method wrapper over the live free function reject_local_disk_symlink_components; no caller in this port (backlog#1823)"
+    )]
    fn reject_symlink_components(&self, path: &Path) -> Result<()> {
        reject_local_disk_symlink_components(self.io_root(), path)
    }

    // Batch path generation with single lock acquisition
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    fn get_object_paths_batch(&self, requests: &[(String, String)]) -> Result<Vec<PathBuf>> {
        let mut results = Vec::with_capacity(requests.len());
        let mut cache_misses = Vec::new();
@@ -6488,12 +6556,13 @@ impl LocalDisk {
        Ok(f)
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    async fn open_file_read_only(&self, path: impl AsRef<Path>) -> Result<File> {
        let f = super::fs::open_file(path.as_ref(), O_RDONLY).await.map_err(to_file_error)?;
        Ok(f)
    }

-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn get_metrics(&self) -> DiskMetrics {
        DiskMetrics::default()
    }
@@ -13,7 +13,6 @@
 // limitations under the License.

 // #730: disk abstractions still carry staged health and direct-I/O migration paths.
-#![allow(dead_code)]

 pub mod disk_store;
 pub mod endpoint;
@@ -1114,6 +1113,10 @@ pub struct DiskInfo {
 }

 #[derive(Clone, Debug, Default)]
+#[allow(
+    dead_code,
+    reason = "MinIO-parity disk info shape with no constructor in this port (backlog#1823)"
+)]
 pub struct Info {
    pub total: u64,
    pub free: u64,
@@ -1372,6 +1375,7 @@ pub fn conv_part_err_to_int(err: &Option<Error>) -> usize {
    }
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn has_part_err(part_errs: &[usize]) -> bool {
    part_errs.iter().any(|err| *err != CHECK_PART_SUCCESS)
 }
@@ -571,6 +571,10 @@ fn regular_files(dir: &Path) -> io::Result<Vec<PathBuf>> {

 /// Fdatasync every regular file directly inside `dir`, then fsync the directory
 /// itself.
+#[allow(
+    dead_code,
+    reason = "reached only through sync_dir_files, whose callers are tests (backlog#1823)"
+)]
 pub fn sync_dir_files_std(dir: impl AsRef<Path>) -> io::Result<()> {
    for entry in std::fs::read_dir(dir.as_ref())? {
        let entry = entry?;
@@ -583,6 +587,7 @@ pub fn sync_dir_files_std(dir: impl AsRef<Path>) -> io::Result<()> {

 /// Async wrapper around [`sync_dir_files_std`]. Large directories flush files
 /// concurrently, bounded both per directory and process-wide.
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub async fn sync_dir_files(dir: impl AsRef<Path>) -> io::Result<()> {
    sync_dir_files_with_limiter(dir, Arc::new(Semaphore::new(MAX_PARALLEL_FILE_SYNCS))).await
 }
@@ -1809,10 +1814,6 @@ impl RenameCommitGuard {
        })
    }

-    pub(crate) fn lock_destination_directory_for_path_access(&self, directory: &Path) -> io::Result<RenameDestinationPathGuard> {
-        self.destination_directory_guard(directory, false)
-    }
-
    pub(crate) fn create_destination_directory_for_path_access(
        &self,
        directory: &Path,
@@ -2858,13 +2859,6 @@ pub async fn os_mkdir_all(dir_path: impl AsRef<Path>, base_dir: impl AsRef<Path>
    Ok(())
 }

-/// Check if a file exists.
-/// Returns true if the file exists, false otherwise.
-#[tracing::instrument(level = "debug", skip_all)]
-pub fn file_exists(path: impl AsRef<Path>) -> bool {
-    std::fs::metadata(path.as_ref()).map(|_| true).unwrap_or(false)
-}
-
 /// Whether an [`io::Error`] means "the directory is not empty".
 ///
 /// POSIX lets `rmdir`/`rename` report a non-empty directory as either
@@ -820,10 +820,263 @@ impl BitrotWriterWrapper {
    }
 }

+// --- startup bitrot self-test (rustfs/backlog#1873, MinIO bitrotSelfTest parity) ---
+//
+// A broken hash implementation (bad SIMD feature combination, platform drift, a
+// key-handling regression) fails silently: every shard reads back "corrupt",
+// heal rewrites data that was fine, and cross-platform clusters disagree about
+// which copy is healthy. The self-test below pins the algorithms the moment a
+// process starts, so a drifted build announces itself instead of quietly
+// rewriting objects. See docs/rustfs-heal-scanner-vs-minio-comprehensive-
+// analysis-2026-08-16.md §6 HS-11.
+
+/// Length of the deterministic self-test payload.
+pub const BITROT_SELF_TEST_PAYLOAD_LEN: usize = 4096;
+
+/// Known-answer digest of [`bitrot_self_test_payload`] under `HighwayHash256S`
+/// (the production default). Pinned so any platform or build where the
+/// implementation drifts fails startup instead of miss-hashing shards.
+const BITROT_SELF_TEST_KAT_HIGHWAY_HASH256S: [u8; 32] = [
+    0xb9, 0x32, 0xa2, 0xaa, 0x4a, 0xb7, 0x33, 0x6a, 0xa3, 0xca, 0x7e, 0x61, 0x9d, 0x86, 0x52, 0x14, 0x6e, 0x7f, 0xd8, 0x9e, 0xea,
+    0x08, 0xd9, 0x8c, 0x33, 0x85, 0x87, 0x19, 0x30, 0xd6, 0xed, 0x06,
+];
+
+/// Known-answer digest of the same payload under `HighwayHash256SLegacy`.
+const BITROT_SELF_TEST_KAT_HIGHWAY_HASH256S_LEGACY: [u8; 32] = [
+    0x98, 0x24, 0x71, 0x4f, 0x16, 0xbb, 0x48, 0x39, 0xed, 0x68, 0xfa, 0x63, 0x5e, 0xd9, 0x07, 0x61, 0xdf, 0x0a, 0xff, 0xcf, 0x7d,
+    0x8c, 0xa8, 0xc7, 0xc0, 0xb6, 0x6f, 0x05, 0xdb, 0xda, 0x5a, 0x22,
+];
+
+/// FIPS 180-2 test vector: SHA-256 of the ASCII string "abc". Unlike the
+/// Highway digests above this one is externally verifiable, so it guards the
+/// whole `HashAlgorithm` plumbing even for readers who distrust pinned
+/// self-computed constants.
+const BITROT_SELF_TEST_KAT_SHA256_ABC: [u8; 32] = [
+    0xba, 0x78, 0x16, 0xbf, 0x8f, 0x01, 0xcf, 0xea, 0x41, 0x41, 0x40, 0xde, 0x5d, 0xae, 0x22, 0x23, 0xb0, 0x03, 0x61, 0xa3, 0x96,
+    0x17, 0x7a, 0x9c, 0xb4, 0x10, 0xff, 0x61, 0xf2, 0x00, 0x15, 0xad,
+];
+
+/// Deterministic self-test payload: xorshift64* from a fixed seed, so every
+/// platform and every run hashes the same 4096 bytes.
+fn bitrot_self_test_payload() -> [u8; BITROT_SELF_TEST_PAYLOAD_LEN] {
+    let mut state = 0x9E37_79B9_7F4A_7C15u64;
+    let mut payload = [0u8; BITROT_SELF_TEST_PAYLOAD_LEN];
+    for byte in payload.iter_mut() {
+        state ^= state >> 12;
+        state ^= state << 25;
+        state ^= state >> 27;
+        *byte = state.wrapping_mul(0x2545_F491_4F6C_DD1D) as u8;
+    }
+    payload
+}
+
+/// Why a bitrot self-test failed.
+#[derive(Debug)]
+pub enum BitrotSelfTestError {
+    /// A known-answer digest mismatched the pinned constant.
+    KnownAnswerMismatch {
+        algorithm: &'static str,
+        got: String,
+        want: String,
+    },
+    /// A freshly encoded shard failed `bitrot_verify`.
+    RoundtripVerify { algorithm: &'static str, detail: String },
+    /// A verified roundtrip read back different bytes than were written.
+    RoundtripReadback { algorithm: &'static str },
+    /// A deliberately tampered shard was not rejected by `bitrot_verify`.
+    TamperNotRejected {
+        algorithm: &'static str,
+        tampered: &'static str,
+    },
+}
+
+impl std::fmt::Display for BitrotSelfTestError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            Self::KnownAnswerMismatch { algorithm, got, want } => {
+                write!(f, "known-answer mismatch for {algorithm}: got {got}, want {want}")
+            }
+            Self::RoundtripVerify { algorithm, detail } => write!(f, "{algorithm} roundtrip shard failed verification: {detail}"),
+            Self::RoundtripReadback { algorithm } => write!(f, "{algorithm} roundtrip read back different bytes"),
+            Self::TamperNotRejected { algorithm, tampered } => {
+                write!(f, "{algorithm} tampered shard ({tampered}) was not rejected")
+            }
+        }
+    }
+}
+
+impl std::error::Error for BitrotSelfTestError {}
+
+fn self_test_hex(bytes: &[u8]) -> String {
+    rustfs_utils::hex(bytes)
+}
+
+// (kept as a named one-liner so every KAT failure site reads the same; the
+// underlying formatter is the shared `rustfs_utils::hex`)
+
+/// Compare a digest against its pinned constant. Split out so a test can drive
+/// it with a wrong constant and prove the mismatch path fires.
+fn bitrot_kat_check(
+    algorithm: &'static str,
+    algo: &HashAlgorithm,
+    payload: &[u8],
+    expected: &[u8; 32],
+) -> Result<(), BitrotSelfTestError> {
+    let digest = algo.hash_encode(payload);
+    let digest = digest.as_ref();
+    if digest.len() != expected.len() || digest != expected.as_slice() {
+        return Err(BitrotSelfTestError::KnownAnswerMismatch {
+            algorithm,
+            got: self_test_hex(digest),
+            want: self_test_hex(expected),
+        });
+    }
+    Ok(())
+}
+
+/// Encode `payload` with `shard_size` blocks, verify it end to end, and read
+/// every block back through `BitrotReader` comparing bytes.
+async fn bitrot_roundtrip_check(
+    algorithm: &'static str,
+    algo: HashAlgorithm,
+    payload: &[u8],
+    shard_size: usize,
+) -> Result<(), BitrotSelfTestError> {
+    let mut writer = BitrotWriter::new(std::io::Cursor::new(Vec::<u8>::new()), shard_size, algo.clone());
+    for chunk in payload.chunks(shard_size) {
+        writer
+            .write(chunk)
+            .await
+            .map_err(|err| BitrotSelfTestError::RoundtripVerify {
+                algorithm,
+                detail: format!("encode failed: {err}"),
+            })?;
+    }
+    let encoded = writer.into_inner().into_inner();
+
+    let on_disk = bitrot_shard_file_size(payload.len(), shard_size, algo.clone());
+    if encoded.len() != on_disk {
+        return Err(BitrotSelfTestError::RoundtripVerify {
+            algorithm,
+            detail: format!("encoded {} bytes, size formula says {on_disk}", encoded.len()),
+        });
+    }
+    bitrot_verify(std::io::Cursor::new(encoded.clone()), on_disk, payload.len(), algo.clone(), shard_size)
+        .await
+        .map_err(|err| BitrotSelfTestError::RoundtripVerify {
+            algorithm,
+            detail: err.to_string(),
+        })?;
+
+    let mut reader = BitrotReader::new(std::io::Cursor::new(encoded), shard_size, algo, false);
+    let mut offset = 0usize;
+    while offset < payload.len() {
+        let want = shard_size.min(payload.len() - offset);
+        let mut buf = vec![0u8; want];
+        let read = reader
+            .read(&mut buf)
+            .await
+            .map_err(|err| BitrotSelfTestError::RoundtripVerify {
+                algorithm,
+                detail: format!("read back failed at offset {offset}: {err}"),
+            })?;
+        if read != want || buf[..read] != payload[offset..offset + read] {
+            return Err(BitrotSelfTestError::RoundtripReadback { algorithm });
+        }
+        offset += read;
+    }
+    Ok(())
+}
+
+/// Flip one byte and require `bitrot_verify` to reject the result.
+async fn bitrot_tamper_check(
+    algorithm: &'static str,
+    algo: HashAlgorithm,
+    payload: &[u8],
+    shard_size: usize,
+    tampered: &'static str,
+    flip_at: usize,
+) -> Result<(), BitrotSelfTestError> {
+    let mut writer = BitrotWriter::new(std::io::Cursor::new(Vec::<u8>::new()), shard_size, algo.clone());
+    for chunk in payload.chunks(shard_size) {
+        writer.write(chunk).await.expect("self-test encode should not fail");
+    }
+    let mut corrupt = writer.into_inner().into_inner();
+    let flip_index = flip_at % corrupt.len();
+    corrupt[flip_index] ^= 0x80;
+
+    let on_disk = bitrot_shard_file_size(payload.len(), shard_size, algo.clone());
+    match bitrot_verify(std::io::Cursor::new(corrupt), on_disk, payload.len(), algo, shard_size).await {
+        // The flipped byte must be rejected as a hash mismatch specifically, not
+        // by any incidental read error: an in-memory cursor cannot fail reads,
+        // so accepting any other failure here would mask a verify path that
+        // errors out before it ever compares hashes.
+        Err(err) if err.to_string().contains("hash mismatch") => Ok(()),
+        Ok(()) => Err(BitrotSelfTestError::TamperNotRejected { algorithm, tampered }),
+        Err(err) => Err(BitrotSelfTestError::RoundtripVerify {
+            algorithm,
+            detail: format!("tampered shard rejected with an unexpected error: {err}"),
+        }),
+    }
+}
+
+/// Verify every bitrot algorithm this crate can write or verify in production:
+/// both streaming Highway variants roundtrip end to end (encode → size formula
+/// → `bitrot_verify` → read back) and reject a flipped byte in both the data
+/// and the leading hash, while all three hashed algorithms reproduce their
+/// pinned known-answer digests.
+///
+/// Runs in well under a millisecond on 4 KiB of data; callers may run it inline
+/// at startup. Pure CPU, no allocation beyond a few KiB of scratch.
+pub async fn bitrot_self_test() -> Result<(), BitrotSelfTestError> {
+    let payload = bitrot_self_test_payload();
+
+    // Externally verifiable vector first: it guards the HashAlgorithm plumbing
+    // itself, before any self-pinned constants are consulted.
+    let abc = HashAlgorithm::SHA256.hash_encode(b"abc");
+    if abc.as_ref() != BITROT_SELF_TEST_KAT_SHA256_ABC.as_slice() {
+        return Err(BitrotSelfTestError::KnownAnswerMismatch {
+            algorithm: "SHA256",
+            got: self_test_hex(abc.as_ref()),
+            want: self_test_hex(&BITROT_SELF_TEST_KAT_SHA256_ABC),
+        });
+    }
+
+    bitrot_kat_check(
+        "HighwayHash256S",
+        &HashAlgorithm::HighwayHash256S,
+        &payload,
+        &BITROT_SELF_TEST_KAT_HIGHWAY_HASH256S,
+    )?;
+    bitrot_kat_check(
+        "HighwayHash256SLegacy",
+        &HashAlgorithm::HighwayHash256SLegacy,
+        &payload,
+        &BITROT_SELF_TEST_KAT_HIGHWAY_HASH256S_LEGACY,
+    )?;
+
+    for (algorithm, algo) in [
+        ("HighwayHash256S", HashAlgorithm::HighwayHash256S),
+        ("HighwayHash256SLegacy", HashAlgorithm::HighwayHash256SLegacy),
+    ] {
+        // Full blocks plus a partial tail, exactly like a real part stripe.
+        let tail_len = 2 * 1024 + 333;
+        bitrot_roundtrip_check(algorithm, algo.clone(), &payload, 1024).await?;
+        bitrot_roundtrip_check(algorithm, algo.clone(), &payload[..tail_len], 1024).await?;
+        // One flipped byte in the final data block, one in the first leading
+        // hash: both must fail verification.
+        bitrot_tamper_check(algorithm, algo.clone(), &payload, 1024, "final data byte", payload.len() - 1).await?;
+        bitrot_tamper_check(algorithm, algo, &payload, 1024, "leading hash byte", 0).await?;
+    }
+
+    Ok(())
+}
+
 #[cfg(test)]
 mod tests {
    use super::{
-        BitrotReader, BitrotWriter, BitrotWriterWrapper, CustomWriter, bitrot_shard_file_size, bitrot_verify, write_all_vectored,
+        BitrotReader, BitrotWriter, BitrotWriterWrapper, CustomWriter, bitrot_kat_check, bitrot_self_test,
+        bitrot_self_test_payload, bitrot_shard_file_size, bitrot_verify, write_all_vectored,
    };
    use super::{MAX_RETAINED_CHUNKS_PER_BLOCK, ShardChunkRead, ShardSource};
    use bytes::Bytes;
@@ -1090,6 +1343,32 @@ mod tests {
        }
    }

+    #[test]
+    fn bitrot_self_test_payload_is_deterministic() {
+        // Two independent builds of the payload must agree byte for byte, or
+        // the pinned known-answer digests below would be meaningless.
+        assert_eq!(bitrot_self_test_payload(), bitrot_self_test_payload());
+    }
+
+    #[test]
+    fn bitrot_self_test_rejects_a_wrong_known_answer_digest() {
+        let payload = bitrot_self_test_payload();
+        let wrong = [0u8; 32];
+        let err = bitrot_kat_check("HighwayHash256S", &HashAlgorithm::HighwayHash256S, &payload, &wrong)
+            .expect_err("a zeroed digest must never match");
+        match err {
+            super::BitrotSelfTestError::KnownAnswerMismatch { algorithm, .. } => assert_eq!(algorithm, "HighwayHash256S"),
+            other => panic!("expected KnownAnswerMismatch, got {other:?}"),
+        }
+    }
+
+    #[tokio::test]
+    async fn bitrot_self_test_passes() {
+        bitrot_self_test()
+            .await
+            .expect("the pinned digests and roundtrip checks must all pass on this platform");
+    }
+
    #[tokio::test]
    async fn vectored_test_writers_cover_fallback_flush_and_shutdown_paths() {
        let mut counting = VectoredCountingWriter::default();
@@ -1189,7 +1468,7 @@ mod tests {
        let last = corrupt.len() - 1;
        corrupt[last] ^= 0x80;
        let err = bitrot_verify(
-            Cursor::new(corrupt),
+            std::io::Cursor::new(corrupt),
            super::bitrot_shard_file_size(data.len(), shard_size, algo.clone()),
            data.len(),
            algo,
@@ -1282,7 +1561,7 @@ mod tests {

    #[tokio::test]
    async fn bitrot_reader_rejects_output_buffers_larger_than_shard_size() {
-        let mut reader = BitrotReader::new(Cursor::new(Vec::<u8>::new()), 4, HashAlgorithm::None, false);
+        let mut reader = BitrotReader::new(std::io::Cursor::new(Vec::<u8>::new()), 4, HashAlgorithm::None, false);
        let mut out = [0u8; 5];
        let err = reader
            .read(&mut out)
@@ -1407,7 +1686,7 @@ mod tests {
            (HashAlgorithm::HighwayHash256, true),
        ] {
            let label = format!("{algo:?}");
-            let writer = Cursor::new(Vec::<u8>::new());
+            let writer = std::io::Cursor::new(Vec::<u8>::new());
            let mut w = BitrotWriter::new(writer, shard_size, algo.clone());
            w.write(&[7u8; 16]).await.unwrap();
            let written = w.into_inner().into_inner();
@@ -1492,7 +1771,7 @@ mod tests {
    }

    async fn encode_one_block(payload: &[u8], shard_size: usize, algo: HashAlgorithm) -> Vec<u8> {
-        let mut w = BitrotWriter::new(Cursor::new(Vec::<u8>::new()), shard_size, algo);
+        let mut w = BitrotWriter::new(std::io::Cursor::new(Vec::<u8>::new()), shard_size, algo);
        w.write(payload).await.unwrap();
        w.into_inner().into_inner()
    }
@@ -1600,7 +1879,7 @@ mod tests {
        for algo in [HashAlgorithm::HighwayHash256S, HashAlgorithm::HighwayHash256SLegacy] {
            for &size in &[1usize, 16, 17, 32, 40, 48] {
                let payload: Vec<u8> = (0..size).map(|i| i as u8).collect();
-                let mut w = BitrotWriter::new(Cursor::new(Vec::<u8>::new()), shard_size, algo.clone());
+                let mut w = BitrotWriter::new(std::io::Cursor::new(Vec::<u8>::new()), shard_size, algo.clone());
                for chunk in payload.chunks(shard_size) {
                    w.write(chunk).await.unwrap();
                }
@@ -1674,14 +1953,14 @@ mod tests {
            w.write(&data).await.expect("write shard");

            let mut via_read = vec![0u8; SHARD];
-            let n1 = BitrotReader::new(Cursor::new(encoded.clone()), SHARD, algo.clone(), false)
+            let n1 = BitrotReader::new(std::io::Cursor::new(encoded.clone()), SHARD, algo.clone(), false)
                .read(&mut via_read)
                .await
                .expect("read");

            // A buffer with only capacity — no initialized bytes at all.
            let mut via_append: Vec<u8> = Vec::with_capacity(SHARD);
-            let n2 = BitrotReader::new(Cursor::new(encoded), SHARD, algo.clone(), false)
+            let n2 = BitrotReader::new(std::io::Cursor::new(encoded), SHARD, algo.clone(), false)
                .read_appending(&mut via_append, SHARD)
                .await
                .expect("read_appending");
@@ -1706,7 +1985,7 @@ mod tests {
            encoded.truncate(encoded.len() - 1);

            let mut out: Vec<u8> = Vec::with_capacity(SHARD);
-            let err = BitrotReader::new(Cursor::new(encoded), SHARD, algo.clone(), false)
+            let err = BitrotReader::new(std::io::Cursor::new(encoded), SHARD, algo.clone(), false)
                .read_appending(&mut out, SHARD)
                .await
                .expect_err("a truncated shard must not succeed");
@@ -1732,7 +2011,7 @@ mod tests {
        encoded[last] ^= 0xff;

        let mut out: Vec<u8> = Vec::with_capacity(SHARD);
-        let err = BitrotReader::new(Cursor::new(encoded), SHARD, algo, false)
+        let err = BitrotReader::new(std::io::Cursor::new(encoded), SHARD, algo, false)
            .read_appending(&mut out, SHARD)
            .await
            .expect_err("a corrupt shard must not verify");
@@ -1844,7 +2123,7 @@ mod tests {
            "Cursor<Bytes> must be able to hand out a block, otherwise the fast path is dead code"
        );
        assert_eq!(mem.position(), 8, "taking a block must advance like a read of the same length");
-        let mut streamed = Cursor::new(encoded.clone());
+        let mut streamed = std::io::Cursor::new(encoded.clone());
        assert!(
            ShardSource::try_take_block(&mut streamed, 8).is_none(),
            "a non-Bytes source must stay on the streaming path"
@@ -1872,7 +2151,7 @@ mod tests {
        );

        let mut via_stream: Vec<u8> = Vec::with_capacity(SHARD);
-        BitrotReader::new(Cursor::new(encoded), SHARD, algo, false)
+        BitrotReader::new(std::io::Cursor::new(encoded), SHARD, algo, false)
            .read_appending(&mut via_stream, SHARD)
            .await
            .expect("streaming read");
@@ -1116,6 +1116,14 @@ mod tests {
        assert!(encoder_source.is::<reed_solomon_erasure::Error>());
    }

+    // The lifecycle transition worker relies on this arm alone to suppress the
+    // closed-connection noise (`bucket_lifecycle_ops.rs`); dropping it here would
+    // silently turn shutdown races back into `error!` log spam.
+    #[test]
+    fn is_network_or_host_down_covers_closed_network_connection() {
+        assert!(is_network_or_host_down("transition failed: use of closed network connection", false));
+    }
+
    // Regression for #952 (ECA-11): an all-`DiskNotFound` slice (every drive in
    // every set unreachable) must NOT be classified as "all not found",
    // otherwise ListObjects silently returns an empty listing and masks a full
@@ -704,6 +704,7 @@ pub(crate) async fn create_bitrot_reader_from_bytes_with_stage_metrics(
 }

 #[allow(clippy::too_many_arguments)]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn create_deferred_bitrot_reader(
    inline_data: Option<Bytes>,
    disk: Option<DiskStore>,
@@ -132,7 +132,6 @@ impl RebalanceStopPropagationRecord {
    }
 }

-#[allow(dead_code)]
 #[derive(Debug, Clone, Default)]
 pub struct DiskStat {
    pub total_space: u64,
@@ -16,8 +16,16 @@ use serde::{Deserialize, Serialize};
 use std::{fmt::Display, io};
 use tracing::info;

+#[allow(
+    dead_code,
+    reason = "tier config wire version stamped by the parity constructors below (backlog#1823)"
+)]
 const C_TIER_CONFIG_VER: &str = "v1";

+#[allow(
+    dead_code,
+    reason = "tier-name validation message reached only from the parity constructors below (backlog#1823)"
+)]
 const ERR_TIER_NAME_EMPTY: &str = "remote tier name empty";
 const WASABI_US_EAST_ENDPOINT: &str = "https://s3.wasabisys.com";
 const WASABI_ALTERNATIVE_ENDPOINTS: &[(&str, &str)] = &[
@@ -264,7 +272,6 @@ impl Clone for TierConfig {
    }
 }

-#[allow(dead_code)]
 impl TierConfig {
    pub(crate) fn clone_with_credentials(&self) -> Self {
        Self {
@@ -284,6 +291,7 @@ impl TierConfig {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn endpoint(&self) -> String {
        match self.tier_type {
            TierType::S3 => self.s3.as_ref().map(|s| s.endpoint.clone()).unwrap_or_default(),
@@ -303,6 +311,7 @@ impl TierConfig {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn bucket(&self) -> String {
        match self.tier_type {
            TierType::S3 => self.s3.as_ref().map(|s| s.bucket.clone()).unwrap_or_default(),
@@ -322,6 +331,7 @@ impl TierConfig {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn prefix(&self) -> String {
        match self.tier_type {
            TierType::S3 => self.s3.as_ref().map(|s| s.prefix.clone()).unwrap_or_default(),
@@ -341,6 +351,7 @@ impl TierConfig {
        }
    }

+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn region(&self) -> String {
        match self.tier_type {
            TierType::S3 => self.s3.as_ref().map(|s| s.region.clone()).unwrap_or_default(),
@@ -457,7 +468,7 @@ impl TierWasabi {
 }

 impl TierS3 {
-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn create<F>(
        name: &str,
        access_key: &str,
@@ -528,7 +539,7 @@ pub struct TierMinIO {
 }

 impl TierMinIO {
-    #[allow(dead_code)]
+    #[allow(dead_code, reason = "MinIO-parity surface with no caller in this port (backlog#1823)")]
    fn create<F>(
        name: &str,
        endpoint: &str,
@@ -14,7 +14,6 @@

 use crate::services::tier::tier::TierConfigMgr;

-#[allow(dead_code)]
 impl TierConfigMgr {
    pub fn msg_size(&self) -> usize {
        100
@@ -180,11 +180,13 @@ pub(in crate::set_disk) enum GetCodecStreamingReaderBuildOutcome {
    Fallback(GetCodecStreamingFallbackReason),
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(in crate::set_disk) struct MultipartCodecStreamingReader {
    pub(in crate::set_disk) readers: VecDeque<Box<dyn AsyncRead + Unpin + Send + Sync>>,
 }

 impl MultipartCodecStreamingReader {
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(in crate::set_disk) fn new(readers: Vec<Box<dyn AsyncRead + Unpin + Send + Sync>>) -> Self {
        Self {
            readers: VecDeque::from(readers),
@@ -666,6 +668,60 @@ pub(in crate::set_disk) async fn data_read_early_stop_inline_body_miss_reason(
    parts_metadata: &[FileInfo],
    disks: &[Option<DiskStore>],
 ) -> Option<&'static str> {
+    if let Some(reason) = data_read_early_stop_inline_candidate_miss_reason(candidate) {
+        return Some(reason);
+    }
+
+    let Ok(erasure) = coding::Erasure::try_new_with_options(
+        candidate.erasure.data_blocks,
+        candidate.erasure.parity_blocks,
+        candidate.erasure.block_size,
+        candidate.uses_legacy_checksum,
+    ) else {
+        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_GEOMETRY);
+    };
+    let data_files =
+        match collect_inline_data_shard_fileinfos_by_index_or_reason(parts_metadata, candidate, erasure.data_shards, |index| {
+            disks.get(index).is_some_and(Option::is_some)
+        }) {
+            Ok(data_files) => data_files,
+            Err(reason) => return Some(reason),
+        };
+
+    let Some(part) = candidate.parts.first() else {
+        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_PART_SHAPE);
+    };
+    let Ok(object_size) = usize::try_from(candidate.size) else {
+        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_SIZE);
+    };
+    let checksum_info = candidate.erasure.get_checksum_info(part.number);
+    let checksum_algo = if candidate.uses_legacy_checksum && checksum_info.algorithm == HashAlgorithm::HighwayHash256S {
+        HashAlgorithm::HighwayHash256SLegacy
+    } else {
+        checksum_info.algorithm
+    };
+    let read_length = inline_erasure_shard_file_offset(
+        0,
+        object_size,
+        object_size,
+        candidate.erasure.block_size,
+        erasure.data_shards,
+        candidate.uses_legacy_checksum,
+    );
+    let shard_size = inline_erasure_shard_size(candidate.erasure.block_size, erasure.data_shards, candidate.uses_legacy_checksum);
+    let Ok(mut readers) =
+        build_inline_bitrot_readers_from_refs(&data_files, bucket, object, read_length, shard_size, &checksum_algo, false).await
+    else {
+        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_BODY_VERIFY);
+    };
+
+    match try_read_inline_data_shards_direct(&mut readers, erasure.data_shards, read_length, object_size).await {
+        Some(body) if body.len() == object_size => None,
+        _ => Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_BODY_VERIFY),
+    }
+}
+
+fn data_read_early_stop_inline_candidate_miss_reason(candidate: &FileInfo) -> Option<&'static str> {
    // `inline_data` excludes remote objects; this diagnostic reports them separately.
    if !rustfs_utils::http::contains_key_str(&candidate.metadata, rustfs_utils::http::SUFFIX_INLINE_DATA) {
        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_NOT_INLINE);
@@ -703,51 +759,7 @@ pub(in crate::set_disk) async fn data_read_early_stop_inline_body_miss_reason(
    if !can_try_inline_data_shards_direct(object_size, candidate.erasure.block_size) {
        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_SIZE);
    }
-
-    let Ok(erasure) = coding::Erasure::try_new_with_options(
-        candidate.erasure.data_blocks,
-        candidate.erasure.parity_blocks,
-        candidate.erasure.block_size,
-        candidate.uses_legacy_checksum,
-    ) else {
-        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_GEOMETRY);
-    };
-    let data_files =
-        match collect_inline_data_shard_fileinfos_by_index_or_reason(parts_metadata, candidate, erasure.data_shards, |index| {
-            disks.get(index).is_some_and(Option::is_some)
-        }) {
-            Ok(data_files) => data_files,
-            Err(reason) => return Some(reason),
-        };
-
-    let Some(part) = candidate.parts.first() else {
-        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_PART_SHAPE);
-    };
-    let checksum_info = candidate.erasure.get_checksum_info(part.number);
-    let checksum_algo = if candidate.uses_legacy_checksum && checksum_info.algorithm == HashAlgorithm::HighwayHash256S {
-        HashAlgorithm::HighwayHash256SLegacy
-    } else {
-        checksum_info.algorithm
-    };
-    let read_length = inline_erasure_shard_file_offset(
-        0,
-        object_size,
-        object_size,
-        candidate.erasure.block_size,
-        erasure.data_shards,
-        candidate.uses_legacy_checksum,
-    );
-    let shard_size = inline_erasure_shard_size(candidate.erasure.block_size, erasure.data_shards, candidate.uses_legacy_checksum);
-    let Ok(mut readers) =
-        build_inline_bitrot_readers_from_refs(&data_files, bucket, object, read_length, shard_size, &checksum_algo, false).await
-    else {
-        return Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_BODY_VERIFY);
-    };
-
-    match try_read_inline_data_shards_direct(&mut readers, erasure.data_shards, read_length, object_size).await {
-        Some(body) if body.len() == object_size => None,
-        _ => Some(GET_METADATA_EARLY_STOP_REASON_DATA_READ_INLINE_BODY_VERIFY),
-    }
+    None
 }

 fn data_read_inline_missing_shards_are_pending(
@@ -1836,6 +1848,7 @@ pub(in crate::set_disk) async fn create_bitrot_readers_until_quorum_all_shards(
 }

 #[allow(clippy::too_many_arguments)]
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(in crate::set_disk) async fn create_bitrot_readers_until_quorum(
    files: &[FileInfo],
    disks: &[Option<DiskStore>],
@@ -2126,6 +2139,7 @@ pub(in crate::set_disk) async fn create_data_block_bitrot_readers(
    setup
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(in crate::set_disk) async fn collect_read_multiple_results<F>(
    tasks: Vec<F>,
    read_quorum: usize,
@@ -2442,6 +2456,7 @@ impl SetDisks {
        let bucket: Arc<str> = Arc::from(bucket);
        let object: Arc<str> = Arc::from(object);
        let version_id: Arc<str> = Arc::from(version_id);
+        let slowtail_fault = get_metadata_slowtail_fault_request(bucket.as_ref(), object.as_ref(), read_data);
        let futures = disks.iter().enumerate().map(|(disk_index, disk)| {
            let disk = disk.clone();
            let task_opts = opts;
@@ -2449,10 +2464,14 @@ impl SetDisks {
            let bucket = bucket.clone();
            let object = object.clone();
            let version_id = version_id.clone();
+            let slowtail_fault = slowtail_fault.clone();
            tokio::spawn(async move {
                let response_start = observe.then(Instant::now);
                let result = if let Some(disk) = disk {
                    Self::record_read_version_call(&object, disk_index);
+                    if let Some(delay) = slowtail_fault.as_ref().and_then(|fault| fault.delay_for_disk(disk_index)) {
+                        tokio::time::sleep(delay).await;
+                    }
                    disk.read_version(&org_bucket, &bucket, &object, &version_id, &task_opts)
                        .await
                } else {
@@ -2548,6 +2567,7 @@ impl SetDisks {
        let mut scheduled_count = 0usize;
        let mut force_full_wait = false;
        let mut final_miss_reason_override = None;
+        let slowtail_fault = get_metadata_slowtail_fault_request(bucket.as_ref(), object.as_ref(), read_data);
        let spawn_read_version =
            |join_set: &mut JoinSet<(usize, disk::error::Result<FileInfo>, Duration)>, index: usize, disk: Option<DiskStore>| {
                let task_opts = opts;
@@ -2555,6 +2575,7 @@ impl SetDisks {
                let bucket = bucket.clone();
                let object = object.clone();
                let version_id = version_id.clone();
+                let slowtail_fault = slowtail_fault.clone();
                join_set.spawn(async move {
                    let response_start = Instant::now();
                    let result = if let Some(disk) = disk {
@@ -2563,6 +2584,9 @@ impl SetDisks {
                        Self::record_read_version_call(&object, index);
                        #[cfg(test)]
                        Self::read_version_fanout_barrier(&object, index).await;
+                        if let Some(delay) = slowtail_fault.as_ref().and_then(|fault| fault.delay_for_disk(index)) {
+                            tokio::time::sleep(delay).await;
+                        }
                        disk.read_version(&org_bucket, &bucket, &object, &version_id, &task_opts)
                            .await
                    } else {
@@ -2596,6 +2620,14 @@ impl SetDisks {
                    Ok(file_info) => {
                        observations.push(MetadataFanoutObservation::from_file_info(&file_info, elapsed));
                        accumulator.observe_file_info(&file_info);
+                        if bounded_fanout
+                            && read_data
+                            && !force_full_wait
+                            && let Some(reason) = data_read_early_stop_inline_candidate_miss_reason(&file_info)
+                        {
+                            force_full_wait = true;
+                            final_miss_reason_override.get_or_insert(reason);
+                        }
                        if let Some(slot) = ress.get_mut(index) {
                            *slot = file_info;
                        }
@@ -2955,6 +2987,7 @@ impl SetDisks {
        (meta_file_infos, errs)
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(in crate::set_disk) async fn read_multiple_files(
        disks: &[Option<DiskStore>],
        req: ReadMultipleReq,
@@ -2988,14 +3021,11 @@ impl SetDisks {
            });
        }

-        let (ress, errors) = match collect_read_multiple_results(futures, read_quorum).await {
+        let (ress, _errors) = match collect_read_multiple_results(futures, read_quorum).await {
            Ok(collected) => collected,
            Err(()) => return empty_quorum_result(),
        };

-        // debug!("ReadMultipleResp ress {:?}", ress);
-        // debug!("ReadMultipleResp errors {:?}", errors);
-
        let mut ret = Vec::with_capacity(req.files.len());

        for want in req.files.iter() {
@@ -3134,6 +3164,7 @@ pub(in crate::set_disk) struct RenameDataCommit {
    pub(in crate::set_disk) committed_file_info: FileInfo,
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 type RenameDataLegacyTuple = (
    Vec<Option<DiskStore>>,
    RenameConvergence,
@@ -3143,6 +3174,7 @@ type RenameDataLegacyTuple = (
 );

 impl RenameDataCommit {
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    fn into_legacy_tuple(self) -> RenameDataLegacyTuple {
        (
            self.online_disks,
@@ -3261,6 +3293,7 @@ impl SetDisks {

    #[tracing::instrument(level = "debug", skip(disks, file_infos))]
    #[allow(clippy::type_complexity)]
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(in crate::set_disk) async fn rename_data(
        disks: &[Option<DiskStore>],
        src_bucket: &str,
@@ -4827,6 +4860,14 @@ impl SetDisks {
    /// is best-effort maintenance: individual delete failures are logged and
    /// skipped rather than propagated.
    pub(crate) async fn reclaim_orphan_data_dirs(&self, bucket: &str, object: &str) -> disk::error::Result<usize> {
+        self.reclaim_orphan_data_dirs_inner(bucket, object, false).await
+    }
+
+    pub(crate) async fn dry_run_reclaim_orphan_data_dirs(&self, bucket: &str, object: &str) -> disk::error::Result<usize> {
+        self.reclaim_orphan_data_dirs_inner(bucket, object, true).await
+    }
+
+    async fn reclaim_orphan_data_dirs_inner(&self, bucket: &str, object: &str, dry_run: bool) -> disk::error::Result<usize> {
        let disks = self.get_disks_internal().await;

        // Phase 1 (read-only): build the referenced-data-dir union and record the
@@ -4934,6 +4975,20 @@ impl SetDisks {
                    continue;
                }
                let stray = format!("{object}/{dir}");
+                if dry_run {
+                    removed += 1;
+                    debug!(
+                        target: "rustfs_ecstore::set_disk",
+                        event = "heal_abandoned_parts",
+                        component = "ecstore",
+                        subsystem = "heal",
+                        state = "dry_run_matched",
+                        result = "matched",
+                        bucket, object, data_dir = %dir,
+                        "Heal abandoned parts dry-run matched orphaned data directory"
+                    );
+                    continue;
+                }
                match disk
                    .delete(
                        bucket,
@@ -5073,6 +5128,7 @@ fn is_cleanup_not_found(e: &DiskError) -> bool {
 /// normalized to `DiskNotFound`: a panic is not a "disk absent" condition and
 /// must not be silently swallowed as an ignorable error (fixes the historical
 /// `Unexpected`/`DiskNotFound` misclassification).
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn map_cleanup_join_result(joined: std::result::Result<Option<DiskError>, tokio::task::JoinError>) -> Option<DiskError> {
    match joined {
        Ok(res) => res,
@@ -5297,6 +5353,7 @@ pub(in crate::set_disk) mod rename_fanout_barrier_phase {
    /// The per-disk old-data-dir cleanup phase of the commit fan-out.
    pub const CLEANUP: &str = "cleanup";
    /// The per-disk `read_version` phase of metadata read fan-out.
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub const READ_VERSION: &str = "read_version";
 }

@@ -5734,6 +5791,130 @@ mod tests {
        (dirs, disks)
    }

+    #[test]
+    fn metadata_slowtail_fault_delay_parses_and_filters_request() {
+        temp_env::with_vars(
+            [
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, Some("25")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS, Some("1,3")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET, Some("bench-bucket")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX, Some("objects/")),
+            ],
+            || {
+                assert_eq!(
+                    get_metadata_slowtail_fault_delay("bench-bucket", "objects/000001", 3, true),
+                    Some(Duration::from_millis(25))
+                );
+                assert!(get_metadata_slowtail_fault_delay("bench-bucket", "objects/000001", 2, true).is_none());
+                assert!(get_metadata_slowtail_fault_delay("other-bucket", "objects/000001", 3, true).is_none());
+                assert!(get_metadata_slowtail_fault_delay("bench-bucket", "other/000001", 3, true).is_none());
+                assert!(get_metadata_slowtail_fault_delay("bench-bucket", "objects/000001", 3, false).is_none());
+            },
+        );
+    }
+
+    #[test]
+    fn metadata_slowtail_fault_delay_disables_invalid_disk_list() {
+        temp_env::with_vars(
+            [
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, Some("25")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS, Some("1,nope")),
+            ],
+            || {
+                assert!(get_metadata_slowtail_fault_delay("bucket", "object", 1, true).is_none());
+            },
+        );
+    }
+
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    async fn metadata_slowtail_fault_delays_only_data_read_metadata_task() {
+        const DISKS: usize = 4;
+        let bucket = "metadata-slowtail-fault-bucket";
+        let object = "objects/metadata-slowtail-fault-object";
+        let (dirs, disks) = call_counter_local_disks(bucket, DISKS).await;
+        install_metadata_fanout_fileinfo(&disks, bucket, object, None).await;
+
+        temp_env::async_with_vars(
+            [
+                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE, Some("false")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, Some("150")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS, Some("3")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET, Some(bucket)),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX, Some("objects/")),
+            ],
+            async {
+                let read_without_data =
+                    SetDisks::read_all_fileinfo_observed(&disks, bucket, bucket, object, "", false, false, false, true, 2);
+                tokio::time::timeout(Duration::from_millis(100), read_without_data)
+                    .await
+                    .expect("non-data metadata fanout must not be delayed by the data-read slowtail hook")
+                    .expect("metadata fanout without read_data should resolve");
+
+                let mut read_with_data = Box::pin(SetDisks::read_all_fileinfo_observed(
+                    &disks, bucket, bucket, object, "", true, false, false, true, 2,
+                ));
+                assert!(
+                    tokio::time::timeout(Duration::from_millis(40), &mut read_with_data)
+                        .await
+                        .is_err(),
+                    "data-read metadata fanout must wait for the injected slow read_version response"
+                );
+                let (parts_metadata, errs, diagnostics) = tokio::time::timeout(Duration::from_secs(2), read_with_data)
+                    .await
+                    .expect("injected slowtail should eventually complete")
+                    .expect("data-read metadata fanout should resolve");
+                assert_eq!(parts_metadata.iter().filter(|fi| fi.name == object).count(), DISKS);
+                assert!(errs.iter().all(Option::is_none));
+                assert_eq!(diagnostics.total_responses(), DISKS);
+            },
+        )
+        .await;
+
+        drop(dirs);
+    }
+
+    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+    async fn metadata_slowtail_fault_delays_early_stop_metadata_task() {
+        const DISKS: usize = 4;
+        let bucket = "metadata-slowtail-early-stop-bucket";
+        let object = "objects/metadata-slowtail-early-stop-object";
+        let (dirs, disks) = call_counter_local_disks(bucket, DISKS).await;
+        install_metadata_fanout_fileinfo(&disks, bucket, object, None).await;
+
+        temp_env::async_with_vars(
+            [
+                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE, Some("true")),
+                (ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE, Some("true")),
+                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT, Some("false")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, Some("150")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS, Some("3")),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET, Some(bucket)),
+                (ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX, Some("objects/")),
+            ],
+            async {
+                let mut read_with_data = Box::pin(SetDisks::read_all_fileinfo_observed(
+                    &disks, bucket, bucket, object, "", true, false, false, true, 2,
+                ));
+                assert!(
+                    tokio::time::timeout(Duration::from_millis(40), &mut read_with_data)
+                        .await
+                        .is_err(),
+                    "early-stop metadata fanout must still wait for the injected slow response after fallback to full wait"
+                );
+                let (parts_metadata, errs, diagnostics) = tokio::time::timeout(Duration::from_secs(2), read_with_data)
+                    .await
+                    .expect("injected early-stop slowtail should eventually complete")
+                    .expect("early-stop metadata fanout should resolve");
+                assert_eq!(parts_metadata.iter().filter(|fi| fi.name == object).count(), DISKS);
+                assert!(errs.iter().all(Option::is_none));
+                assert_eq!(diagnostics.total_responses(), DISKS);
+            },
+        )
+        .await;
+
+        drop(dirs);
+    }
+
    /// Demo / regression guard for the backlog#1325 per-disk call counters.
    ///
    /// The metadata fan-out issues each `read_version` inside its own
@@ -7081,7 +7262,7 @@ mod tests {
    }

    #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-    async fn bounded_non_inline_data_get_hedges_then_waits_for_full_fanout() {
+    async fn bounded_non_inline_data_get_immediately_forces_full_fanout() {
        const DISKS: usize = 4;
        let bucket = "bounded-data-get-hedge-bucket";
        let object = "bounded-data-get-hedge-object";
@@ -7112,7 +7293,7 @@ mod tests {
                    }
                })
                .await
-                .expect("bounded data-read fanout should hedge by starting the spare disk");
+                .expect("bounded non-inline data-read fanout should immediately schedule the spare disk");

                let pending = tokio::time::timeout(BARRIER_PAUSE_GUARD, &mut read).await;
                assert!(
@@ -7128,7 +7309,7 @@ mod tests {
                assert_eq!(
                    calls.total(disk_call_counters::KIND_READ_VERSION),
                    DISKS as u64,
-                    "bounded data-read fanout should issue the paused disk plus one spare hedge"
+                    "bounded non-inline data-read fanout should issue the paused disk plus the remaining spare"
                );
                assert_eq!(diagnostics.total_responses(), DISKS);
                assert_eq!(parts_metadata.iter().filter(|fi| fi.name == object).count(), DISKS);
@@ -7155,16 +7336,42 @@ mod tests {
                ("RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT", None::<&str>),
            ],
            async {
+                let barrier = rename_fanout_barrier::arm(object, 2, rename_fanout_barrier::PHASE_READ_VERSION);
                let calls = disk_call_counters::observe(object);
-                let (parts_metadata, errs, diagnostics) =
-                    SetDisks::read_all_fileinfo_observed(&disks, bucket, bucket, object, "", true, false, false, true, 2)
+                let disks_for_read = disks.clone();
+                let mut read = tokio::spawn(async move {
+                    SetDisks::read_all_fileinfo_observed(&disks_for_read, bucket, bucket, object, "", true, false, false, true, 2)
                        .await
-                        .expect("default data-read metadata should resolve");
+                });
+
+                tokio::time::timeout(BARRIER_PAUSE_GUARD, barrier.wait_until_paused())
+                    .await
+                    .expect("default bounded non-inline read should schedule the paused metadata task");
+                tokio::time::timeout(BARRIER_PAUSE_GUARD, async {
+                    while calls.for_disk(disk_call_counters::KIND_READ_VERSION, 3) == 0 {
+                        tokio::task::yield_now().await;
+                    }
+                })
+                .await
+                .expect(
+                    "default bounded non-inline read should immediately force full fanout after the first non-inline response",
+                );
+
+                let pending = tokio::time::timeout(BARRIER_PAUSE_GUARD, &mut read).await;
+                assert!(
+                    pending.is_err(),
+                    "default non-inline data reads must not return before the paused metadata response"
+                );
+                barrier.release();
+                let (parts_metadata, errs, diagnostics) = read
+                    .await
+                    .expect("metadata read task should not panic")
+                    .expect("default data-read metadata should resolve");

                assert_eq!(
                    calls.total(disk_call_counters::KIND_READ_VERSION),
                    DISKS as u64,
-                    "default non-inline GET data-read metadata must keep full fanout for read-failure tolerance"
+                    "default non-inline GET data-read metadata must keep full fanout without waiting for a quorum miss first"
                );
                assert_eq!(diagnostics.total_responses(), DISKS);
                assert_eq!(parts_metadata.iter().filter(|fi| fi.name == object).count(), DISKS);
@@ -42,12 +42,20 @@ impl<'a> SetDisksCtx<'a> {
    }

    /// The borrowed core, for state not yet fronted by a typed accessor.
+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn core(&self) -> &'a SetDisks {
        self.core
    }

    // --- Immutable topology / config (fixed after construction) ---

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn set_index(&self) -> usize {
        self.core.set_index
    }
@@ -56,14 +64,26 @@ impl<'a> SetDisksCtx<'a> {
        self.core.pool_index
    }

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn set_drive_count(&self) -> usize {
        self.core.set_drive_count
    }

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn default_parity_count(&self) -> usize {
        self.core.default_parity_count
    }

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn set_endpoints(&self) -> &'a [Endpoint] {
        &self.core.set_endpoints
    }
@@ -72,6 +92,10 @@ impl<'a> SetDisksCtx<'a> {
        &self.core.format
    }

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn locker_owner(&self) -> &'a str {
        &self.core.locker_owner
    }
@@ -84,6 +108,10 @@ impl<'a> SetDisksCtx<'a> {

    // --- Locker trio ---

+    #[allow(
+        dead_code,
+        reason = "SetDisks split seam (backlog#815) with no caller in this port (backlog#1823)"
+    )]
    pub(crate) fn lockers(&self) -> &'a [Arc<dyn LockClient>] {
        &self.core.lockers
    }
@@ -39,7 +39,6 @@
 //! - `metadata.rs`, `replication.rs`, `shard_source.rs` — supporting helpers.

 // #730: SetDisks still hosts staged read/heal/write migration helpers.
-#![allow(dead_code)]
 #![allow(unused_imports)]
 #![allow(unused_variables)]

@@ -103,9 +102,7 @@ use crate::storage_api_contracts::{
 };
 use crate::store::utils::is_reserved_or_invalid_bucket;
 use crate::{
-    bucket::lifecycle::bucket_lifecycle_ops::{
-        LifecycleOps, gen_transition_objname, get_transitioned_object_reader_with_tier_manager, put_restore_opts,
-    },
+    bucket::lifecycle::bucket_lifecycle_ops::{LifecycleOps, get_transitioned_object_reader_with_tier_manager, put_restore_opts},
    cache_value::metacache_set::{ListPathRawOptions, list_path_raw},
    config::storageclass,
    disk::{
@@ -177,15 +174,14 @@ use std::future::Future;
 use std::hash::{BuildHasher, Hash, Hasher};
 use std::mem::{self};
 use std::pin::Pin;
-use std::sync::OnceLock;
 use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
+use std::sync::{Arc, OnceLock};
 use std::task::{Context, Poll};
 use std::time::{Instant, SystemTime, UNIX_EPOCH};
 use std::{
    collections::{HashMap, HashSet},
    io::{Cursor, Write},
    path::Path,
-    sync::Arc,
    time::Duration,
 };
 use time::OffsetDateTime;
@@ -624,7 +620,9 @@ fn adaptive_duplex_buffer_size(object_size: i64) -> usize {
 // Each flag has a corresponding `*_ROLLOUT_PCT` for percentage-based gradual rollout.
 // ============================================================================

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 const DISK_ONLINE_TIMEOUT: Duration = Duration::from_secs(1);
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 const DISK_HEALTH_CACHE_TTL: Duration = Duration::from_millis(750);
 const GET_OBJECT_METADATA_CACHE_TTL: Duration = Duration::from_secs(2); // Increased from 250ms to 2s
 const DEFAULT_GET_OBJECT_METADATA_CACHE_MAX_ENTRIES: usize = 4096; // Increased from 1024 to 4096
@@ -698,7 +696,15 @@ const ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_METADATA_EAR
 // the env var to `false` to fall back to full-wait metadata fanout.
 const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE: bool = true;

+#[allow(
+    dead_code,
+    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
+)]
 const ENV_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT: &str = "RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT";
+#[allow(
+    dead_code,
+    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
+)]
 const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_ROLLOUT_PCT: u32 = 100;

 const ENV_RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_METADATA_VERSION_EARLY_STOP_ENABLE";
@@ -708,7 +714,12 @@ const ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE: &str = "RUSTFS_GET_ME
 const DEFAULT_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE: bool = true;

 const ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT: &str = "RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT";
-const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT: bool = false;
+const DEFAULT_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT: bool = true;
+
+const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS";
+const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS";
+const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET";
+const ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX: &str = "RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX";

 // --- Multipart Reader-Setup Prefetch Configuration (backlog#870) ---

@@ -910,6 +921,10 @@ mod prepared_get_object_metadata_tests {
            .expect("test should find an object whose initial fanout covers both data shards")
    }

+    #[allow(
+        dead_code,
+        reason = "test fixture no assertion in this module uses today; the live namesake lives in io_primitives tests (backlog#1823)"
+    )]
    fn bounded_spare_disk_index(bucket: &str, object: &str) -> usize {
        *bounded_metadata_fanout_order(bucket, object, 4, 2)
            .get(3)
@@ -1110,7 +1125,10 @@ mod prepared_get_object_metadata_tests {

        assert_eq!(object_size, payload.len() as i64);
        assert_eq!(restored, payload);
-        assert_eq!(calls_total, 4, "default production GET should eagerly schedule the full metadata fanout");
+        assert_eq!(
+            calls_total, 4,
+            "default production inline GET should schedule the initial bounded quorum plus one hedge"
+        );
        assert_eq!(
            recorder.histogram_values(
                "rustfs_io_get_object_metadata_fanout_scheduled",
@@ -1684,6 +1702,95 @@ fn is_get_metadata_early_stop_bounded_fanout_enabled() -> bool {
    }
 }

+#[derive(Debug)]
+struct GetMetadataSlowtailFaultConfig {
+    delay: Duration,
+    disks: Arc<[usize]>,
+    bucket: Option<String>,
+    object_prefix: Option<String>,
+}
+
+#[derive(Clone, Debug)]
+struct GetMetadataSlowtailFaultRequest {
+    delay: Duration,
+    disks: Arc<[usize]>,
+}
+
+impl GetMetadataSlowtailFaultRequest {
+    fn delay_for_disk(&self, disk_index: usize) -> Option<Duration> {
+        self.disks.contains(&disk_index).then_some(self.delay)
+    }
+}
+
+fn parse_get_metadata_slowtail_fault_disks(raw: &str) -> Option<Vec<usize>> {
+    let mut disks = Vec::new();
+    for item in raw.split(',').map(str::trim).filter(|item| !item.is_empty()) {
+        let Ok(index) = item.parse::<usize>() else {
+            return None;
+        };
+        if !disks.contains(&index) {
+            disks.push(index);
+        }
+    }
+    (!disks.is_empty()).then_some(disks)
+}
+
+fn load_get_metadata_slowtail_fault_config() -> Option<GetMetadataSlowtailFaultConfig> {
+    let delay_ms = rustfs_utils::get_env_u64(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DELAY_MS, 0);
+    if delay_ms == 0 {
+        return None;
+    }
+    let disks = parse_get_metadata_slowtail_fault_disks(&std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_DISKS).ok()?)?;
+    let bucket = std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_BUCKET)
+        .ok()
+        .filter(|value| !value.is_empty());
+    let object_prefix = std::env::var(ENV_RUSTFS_GET_METADATA_SLOWTAIL_FAULT_OBJECT_PREFIX)
+        .ok()
+        .filter(|value| !value.is_empty());
+    Some(GetMetadataSlowtailFaultConfig {
+        delay: Duration::from_millis(delay_ms),
+        disks: Arc::from(disks.into_boxed_slice()),
+        bucket,
+        object_prefix,
+    })
+}
+
+fn get_metadata_slowtail_fault_request(bucket: &str, object: &str, read_data: bool) -> Option<GetMetadataSlowtailFaultRequest> {
+    if !read_data {
+        return None;
+    }
+
+    #[cfg(test)]
+    let config = load_get_metadata_slowtail_fault_config();
+    #[cfg(test)]
+    let config = config.as_ref()?;
+    #[cfg(not(test))]
+    let config = ({
+        static CACHED: OnceLock<Option<GetMetadataSlowtailFaultConfig>> = OnceLock::new();
+        CACHED.get_or_init(load_get_metadata_slowtail_fault_config).as_ref()
+    })?;
+
+    if let Some(expected_bucket) = &config.bucket
+        && expected_bucket != bucket
+    {
+        return None;
+    }
+    if let Some(expected_prefix) = &config.object_prefix
+        && !object.starts_with(expected_prefix)
+    {
+        return None;
+    }
+    Some(GetMetadataSlowtailFaultRequest {
+        delay: config.delay,
+        disks: config.disks.clone(),
+    })
+}
+
+#[cfg(test)]
+fn get_metadata_slowtail_fault_delay(bucket: &str, object: &str, disk_index: usize, read_data: bool) -> Option<Duration> {
+    get_metadata_slowtail_fault_request(bucket, object, read_data)?.delay_for_disk(disk_index)
+}
+
 /// Check if multipart reads prefetch the next part's bitrot reader setup
 /// while the current part decodes (backlog#870).
 ///
@@ -1709,6 +1816,10 @@ fn is_multipart_reader_setup_prefetch_enabled() -> bool {
    }
 }

+#[allow(
+    dead_code,
+    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
+)]
 fn get_metadata_early_stop_rollout_pct() -> u32 {
    static CACHED: OnceLock<u32> = OnceLock::new();
    *CACHED.get_or_init(|| {
@@ -1748,6 +1859,10 @@ fn should_use_codec_streaming(config: GetCodecStreamingConfig, bucket: &str, obj
 }

 /// Should this specific request use metadata early-stop?
+#[allow(
+    dead_code,
+    reason = "percentage-rollout facet of the metadata early-stop switch; its predicate has no caller while the sibling enable flag is live (backlog#1823)"
+)]
 pub fn should_use_metadata_early_stop(bucket: &str, object: &str) -> bool {
    let base = is_get_metadata_early_stop_enabled();
    let pct = get_metadata_early_stop_rollout_pct();
@@ -2181,6 +2296,7 @@ fn classify_get_codec_streaming_object_class(
    GetCodecStreamingObjectClass::PlainSinglePart
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn is_get_small_object_direct_memory_eligible_with_threshold(
    range: &Option<HTTPRangeSpec>,
    object_info: &ObjectInfo,
@@ -2786,6 +2902,7 @@ pub struct SetDisks {
    /// Stable namespace shared by every object lock created for this set.
    set_lock_namespace: Arc<str>,
    pub format: FormatV3,
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    disk_health_cache: Arc<RwLock<Vec<Option<DiskHealthEntry>>>>,
    get_object_metadata_cache: moka::future::Cache<GetObjectMetadataCacheKey, Arc<GetObjectMetadataCacheEntry>>,
    get_object_metadata_cache_hash_builder: std::collections::hash_map::RandomState,
@@ -3061,11 +3178,13 @@ struct GetObjectMetadataCacheEntry {

 #[derive(Clone, Debug)]
 struct DiskHealthEntry {
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    last_check: Instant,
    online: bool,
 }

 impl DiskHealthEntry {
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    fn cached_value(&self) -> Option<bool> {
        if self.last_check.elapsed() <= DISK_HEALTH_CACHE_TTL {
            Some(self.online)
@@ -3659,6 +3778,7 @@ fn multipart_put_large_batch_min_size_bytes() -> usize {
    })
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn classify_small_write_path(is_inline_buffer: bool, object_size: i64, block_size: usize) -> SmallWritePath {
    if should_use_inline_small_fast_path(is_inline_buffer, object_size, block_size) {
        SmallWritePath::Inline
@@ -4237,6 +4357,7 @@ fn check_object_lock_retention_update(bucket: &str, object: &str, obj_info: &Obj
 ///
 /// Fail closed: when bucket metadata cannot be resolved the check stays on, so
 /// object-lock protection is never skipped because of a metadata lookup miss.
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub(crate) fn object_lock_delete_check_required(bucket_meta: Option<&crate::bucket::metadata::BucketMetadata>) -> bool {
    bucket_meta.is_none_or(|meta| meta.object_locking())
 }
@@ -4512,15 +4633,6 @@ impl Hash for ObjProps {
    }
 }

-#[derive(Default, Clone, Debug)]
-pub struct HealEntryResult {
-    pub bytes: usize,
-    pub success: bool,
-    pub skipped: bool,
-    pub entry_done: bool,
-    pub name: String,
-}
-
 fn is_object_dangling(
    meta_arr: &[FileInfo],
    errs: &[Option<DiskError>],
@@ -5297,6 +5409,7 @@ pub fn is_valid_storage_class(storage_class: &str) -> bool {
 }

 /// Returns true if the storage class is a cold storage tier that requires special handling
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn is_cold_storage_class(storage_class: &str) -> bool {
    matches!(
        storage_class,
@@ -5305,6 +5418,7 @@ pub fn is_cold_storage_class(storage_class: &str) -> bool {
 }

 /// Returns true if the storage class is an infrequent access tier
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 pub fn is_infrequent_access_class(storage_class: &str) -> bool {
    matches!(
        storage_class,
@@ -6884,6 +6998,100 @@ mod tests {
        assert!(object_dir.join(STORAGE_FORMAT_FILE).exists(), "metadata must be preserved");
    }

+    async fn recv_abandoned_parts_trace(
+        trace: &mut rustfs_common::trace_bus::TraceSubscription,
+        bucket: &str,
+        object: &str,
+        state: &str,
+    ) -> rustfs_common::trace_bus::TraceEvent {
+        for _ in 0..32 {
+            let event = tokio::time::timeout(std::time::Duration::from_secs(1), trace.recv())
+                .await
+                .expect("abandoned-parts trace event should arrive")
+                .expect("trace bus should stay open");
+            if event.kind == rustfs_common::trace_bus::TraceKind::Heal
+                && event.func == rustfs_common::trace_bus::TraceFunc::HealCheckAbandonedParts
+                && event.bucket.as_deref() == Some(bucket)
+                && event.object.as_deref() == Some(object)
+                && trace_attr_string(&event, "state").as_deref() == Some(state)
+            {
+                return (*event).clone();
+            }
+        }
+
+        panic!("expected abandoned-parts trace state {state} for {bucket}/{object}");
+    }
+
+    fn trace_attr_string(event: &rustfs_common::trace_bus::TraceEvent, key: &str) -> Option<String> {
+        event.attrs.iter().find_map(|attr| {
+            if attr.key != key {
+                return None;
+            }
+            Some(match &attr.value {
+                rustfs_common::trace_bus::TraceVal::Bool(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::U64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::I64(value) => value.to_string(),
+                rustfs_common::trace_bus::TraceVal::Str(value) => value.to_string(),
+            })
+        })
+    }
+
+    #[tokio::test]
+    async fn check_abandoned_parts_dry_run_counts_without_deleting() {
+        let mut trace = rustfs_common::trace_bus::subscribe_trace_events();
+        let (dir, disk) = make_single_local_disk().await;
+        let live = Uuid::new_v4();
+        let orphan = Uuid::new_v4();
+
+        let object_dir = dir.path().join("bucket").join("obj");
+        write_object_meta_with_data_dirs(&object_dir, "bucket", "obj", &[live]).await;
+        fs::create_dir_all(object_dir.join(live.to_string()))
+            .await
+            .expect("live data dir should be created");
+        fs::create_dir_all(object_dir.join(orphan.to_string()))
+            .await
+            .expect("orphan data dir should be created");
+
+        let set = make_set_disks_with(vec![Some(disk)]).await;
+        set.check_abandoned_parts(
+            "bucket",
+            "obj",
+            &HealOpts {
+                dry_run: true,
+                no_lock: true,
+                ..Default::default()
+            },
+        )
+        .await
+        .expect("dry-run abandoned-parts check should succeed");
+        let dry_run_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "dry_run_matched").await;
+        assert_eq!(trace_attr_string(&dry_run_trace, "dry_run").as_deref(), Some("true"));
+        assert_eq!(trace_attr_string(&dry_run_trace, "data_dirs").as_deref(), Some("1"));
+
+        assert!(object_dir.join(live.to_string()).exists(), "referenced data dir must be preserved");
+        assert!(object_dir.join(orphan.to_string()).exists(), "dry-run must not remove orphaned data dir");
+
+        set.check_abandoned_parts(
+            "bucket",
+            "obj",
+            &HealOpts {
+                no_lock: true,
+                ..Default::default()
+            },
+        )
+        .await
+        .expect("abandoned-parts check should reclaim stale data dir");
+        let reclaim_trace = recv_abandoned_parts_trace(&mut trace, "bucket", "obj", "reclaimed").await;
+        assert_eq!(trace_attr_string(&reclaim_trace, "dry_run").as_deref(), Some("false"));
+        assert_eq!(trace_attr_string(&reclaim_trace, "data_dirs").as_deref(), Some("1"));
+
+        assert!(
+            object_dir.join(live.to_string()).exists(),
+            "referenced data dir must remain after reclaim"
+        );
+        assert!(!object_dir.join(orphan.to_string()).exists(), "orphaned data dir must be removed");
+    }
+
    #[tokio::test]
    async fn reclaim_orphan_data_dirs_recovers_deferred_cleanup_after_restart() {
        let (dir, disk) = make_single_local_disk().await;
@@ -12119,11 +12327,18 @@ mod tests {
            .expect_err("unsupported copy_object_part should return a typed error");
        assert!(matches!(copy_part_err, StorageError::NotImplemented));

-        let abandoned_err = set_disks
-            .check_abandoned_parts("bucket", "object", &HealOpts::default())
+        set_disks
+            .check_abandoned_parts(
+                "bucket",
+                "object",
+                &HealOpts {
+                    dry_run: true,
+                    no_lock: true,
+                    ..Default::default()
+                },
+            )
            .await
-            .expect_err("abandoned-parts check should stay in the upper reconciliation layer");
-        assert!(matches!(abandoned_err, StorageError::NotImplemented));
+            .expect("abandoned-parts check should be callable on empty disk sets");
    }

    #[tokio::test]
@@ -16,6 +16,7 @@ use super::super::*;
 use crate::disk::disk_store::DiskStoreRenameDataExt;
 use crate::io_support::bitrot::object_mmap_read_enabled;
 use crate::storage_api_contracts::namespace::NamespaceLocking as _;
+use rustfs_common::trace_bus::{TraceEvent, TraceFunc, TraceKind, trace_emit};
 use tracing::trace;

 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
@@ -453,7 +454,9 @@ impl SetDisks {
            ..Default::default()
        };

-        let write_lock_guard = if !opts.no_lock {
+        // Bound, not `_`: this guard must live to the end of the scope. A bare
+        // `_` would drop it here and release the namespace write lock.
+        let _write_lock_guard = if !opts.no_lock {
            let ns_lock = self.new_ns_lock(bucket, object).await?;
            Some(
                ns_lock
@@ -996,7 +999,7 @@ impl SetDisks {
                                                readers.push(None);
                                                continue;
                                            }
-                                            Err(e) => {
+                                            Err(_e) => {
                                                readers.push(None);
                                                continue;
                                            }
@@ -1545,6 +1548,9 @@ impl SetDisks {

        for candidate in candidates.iter_mut().filter(|candidate| candidate.local_payload) {
            for (disk_index, disk) in disks.iter().enumerate() {
+                // Only the #[cfg(test)] fault-injection branch below reads this.
+                #[cfg(not(test))]
+                let _ = disk_index;
                let Some(disk) = disk else {
                    return Ok(DanglingDeleteSafety::UnsafeToDelete);
                };
@@ -1716,6 +1722,10 @@ impl SetDisks {
        Ok((result, None))
    }

+    #[allow(
+        dead_code,
+        reason = "lock-taking wrapper over the live heal_object_dir_locked; only comments reference it (backlog#1823)"
+    )]
    #[tracing::instrument(level = "trace", skip(self), fields(bucket = %bucket, object = %object))]
    pub(in crate::set_disk) async fn heal_object_dir(
        &self,
@@ -2048,11 +2058,61 @@ impl crate::storage_api_contracts::heal::HealOperations for SetDisks {
        Err(Error::DiskNotFound)
    }

-    #[tracing::instrument(skip(self))]
-    async fn check_abandoned_parts(&self, _bucket: &str, _object: &str, _opts: &HealOpts) -> Result<()> {
-        // Multipart orphan reconciliation is intentionally retained above the set layer
-        // until there is a concrete caller and a stable lower-level contract to implement.
-        Err(StorageError::NotImplemented)
+    #[tracing::instrument(level = "debug", skip(self, opts), fields(bucket = %bucket, object = %object, dry_run = opts.dry_run))]
+    async fn check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
+        let started_at = std::time::Instant::now();
+        let _write_lock_guard = if !opts.no_lock {
+            let ns_lock = self.new_ns_lock(bucket, object).await?;
+            Some(
+                ns_lock
+                    .get_write_lock(get_lock_acquire_timeout())
+                    .await
+                    .map_err(|e| self.map_namespace_lock_error(bucket, object, "write", e))?,
+            )
+        } else {
+            None
+        };
+
+        let removed = if opts.dry_run {
+            self.dry_run_reclaim_orphan_data_dirs(bucket, object).await?
+        } else {
+            self.reclaim_orphan_data_dirs(bucket, object).await?
+        };
+        let state = if opts.dry_run && removed > 0 {
+            "dry_run_matched"
+        } else if removed > 0 {
+            "reclaimed"
+        } else {
+            "checked"
+        };
+        let data_dirs = u64::try_from(removed).unwrap_or(u64::MAX);
+
+        trace_emit(|| {
+            TraceEvent::new(TraceKind::Heal, TraceFunc::HealCheckAbandonedParts)
+                .with_bucket(bucket)
+                .with_object(object)
+                .with_duration(started_at.elapsed())
+                .with_attr("state", state)
+                .with_attr("dry_run", opts.dry_run)
+                .with_attr("data_dirs", data_dirs)
+        });
+
+        if removed > 0 {
+            trace!(
+                event = "heal_abandoned_parts",
+                component = LOG_COMPONENT_ECSTORE,
+                subsystem = LOG_SUBSYSTEM_HEAL,
+                state = if opts.dry_run { "dry_run_matched" } else { "reclaimed" },
+                result = "ok",
+                bucket,
+                object,
+                dry_run = opts.dry_run,
+                data_dirs = removed,
+                "Heal abandoned parts checked object data directories"
+            );
+        }
+
+        Ok(())
    }
 }

@@ -3237,4 +3297,223 @@ mod heal_result_report_tests {
        assert!(result.detail.contains("part 1"));
        assert!(result.detail.contains("bitrot_failure=true"));
    }
+
+    // HS-12 (backlog#1874): a versioned DELETE racing an object heal must never
+    // resurrect the deleted version. The heal has real reconstruction work (a
+    // shard of the doomed version is removed), so both sides touch the same
+    // (bucket, object, data_dir); whichever order the ns write lock serializes
+    // them in, the committed delete must win.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn heal_racing_version_delete_never_resurrects_the_deleted_version() {
+        let (temp_dirs, disks, set) = hermetic_set_disks_isolated(4).await;
+        let bucket = "heal-race-delete-no-resurrect";
+        let object = "object.bin";
+        set.make_bucket(
+            bucket,
+            &MakeBucketOptions {
+                versioning_enabled: true,
+                ..Default::default()
+            },
+        )
+        .await
+        .expect("versioned bucket should be created");
+
+        let mut first_reader = PutObjReader::from_vec(vec![0x11; 1024 * 1024]);
+        let first_info = set
+            .put_object(
+                bucket,
+                object,
+                &mut first_reader,
+                &ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("first version should be written");
+        let first_version = first_info
+            .version_id
+            .expect("versioned put should return the first version id")
+            .to_string();
+
+        let mut second_reader = PutObjReader::from_vec(vec![0x22; 1024 * 1024]);
+        let second_info = set
+            .put_object(
+                bucket,
+                object,
+                &mut second_reader,
+                &ObjectOptions {
+                    versioned: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("second version should be written");
+        let second_version = second_info
+            .version_id
+            .expect("versioned put should return the second version id")
+            .to_string();
+
+        // Damage one shard of the doomed version so the racing heal performs an
+        // actual reconstruction over its data dir instead of an early exit.
+        let doomed_source = disks[0]
+            .read_version("", bucket, object, &first_version, &ReadOptions::default())
+            .await
+            .expect("doomed version metadata should be readable");
+        let doomed_data_dir = doomed_source
+            .data_dir
+            .expect("non-inline version should have a data directory");
+        tokio::fs::remove_file(
+            temp_dirs[1]
+                .path()
+                .join(bucket)
+                .join(object)
+                .join(doomed_data_dir.to_string())
+                .join("part.1"),
+        )
+        .await
+        .expect("shard damage should be injected before the race");
+
+        let delete_set = set.clone();
+        let (delete_res, heal_res) = tokio::join!(
+            async {
+                delete_set
+                    .delete_object(
+                        bucket,
+                        object,
+                        ObjectOptions {
+                            versioned: true,
+                            version_id: Some(first_version.clone()),
+                            object_lock_config_snapshot: Some(Arc::new(crate::set_disk::ObjectLockConfigSnapshot::new(
+                                crate::bucket::metadata_sys::ObjectLockConfigState::ConfirmedAbsent,
+                            ))),
+                            ..Default::default()
+                        },
+                    )
+                    .await
+            },
+            async {
+                set.heal_object(
+                    bucket,
+                    object,
+                    "",
+                    &HealOpts {
+                        scan_mode: HealScanMode::Deep,
+                        ..Default::default()
+                    },
+                )
+                .await
+            },
+        );
+        delete_res.expect("version delete must succeed under lock serialization");
+        // The heal may legitimately report a transient failure when the version
+        // it was rebuilding disappears mid-flight; only the end state matters.
+        drop(heal_res);
+
+        let resurrected = set
+            .get_object_info(
+                bucket,
+                object,
+                &ObjectOptions {
+                    versioned: true,
+                    version_id: Some(first_version.clone()),
+                    ..Default::default()
+                },
+            )
+            .await;
+        assert!(
+            matches!(&resurrected, Err(Error::FileVersionNotFound) | Err(Error::ObjectNotFound(..))),
+            "a racing heal must not resurrect the deleted version: {resurrected:?}"
+        );
+
+        let survivor = set
+            .get_object_info(
+                bucket,
+                object,
+                &ObjectOptions {
+                    versioned: true,
+                    version_id: Some(second_version.clone()),
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("surviving version must remain readable after the race");
+        assert_eq!(survivor.size, 1024 * 1024, "survivor size must be intact");
+    }
+
+    // HS-12 (backlog#1874): unversioned overwrite commits race a Deep heal on
+    // the same object. The overwrite's post-commit tail deletes the replaced
+    // data dir without the ns lock (object.rs commit tail), which is exactly
+    // the intersection the audit flagged: the heal must tolerate the tail race
+    // (retryable outcome) and every committed overwrite must survive — the
+    // final current version is exactly the last payload written.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn heal_racing_unversioned_overwrites_preserves_the_last_commit() {
+        let (temp_dirs, disks, set) = hermetic_set_disks_isolated(4).await;
+        let bucket = "heal-race-put-overwrite";
+        let object = "object.bin";
+        set.make_bucket(bucket, &MakeBucketOptions::default())
+            .await
+            .expect("bucket should be created");
+
+        const ROUNDS: usize = 8;
+        const PAYLOAD_SIZE: usize = 256 * 1024;
+        let mut last_etag = String::new();
+        for round in 0..ROUNDS {
+            // Give the heal something to rebuild on alternating rounds: remove a
+            // shard of the current data dir right before the race.
+            if round % 2 == 1 {
+                let current = disks[2]
+                    .read_version("", bucket, object, "", &ReadOptions::default())
+                    .await
+                    .expect("current metadata should be readable");
+                if let Some(data_dir) = current.data_dir {
+                    let shard = temp_dirs[3]
+                        .path()
+                        .join(bucket)
+                        .join(object)
+                        .join(data_dir.to_string())
+                        .join("part.1");
+                    if shard.exists() {
+                        tokio::fs::remove_file(&shard)
+                            .await
+                            .expect("shard damage should be injectable mid-race");
+                    }
+                }
+            }
+
+            let payload = vec![round as u8; PAYLOAD_SIZE];
+            let mut put_reader = PutObjReader::from_vec(payload);
+            let put_opts = ObjectOptions::default();
+            let heal_opts = HealOpts {
+                scan_mode: HealScanMode::Deep,
+                ..Default::default()
+            };
+            let (put_res, heal_res) = tokio::join!(
+                set.put_object(bucket, object, &mut put_reader, &put_opts),
+                set.heal_object(bucket, object, "", &heal_opts),
+            );
+            let put_info = put_res.expect("overwrite must succeed under lock serialization");
+            last_etag = put_info.etag.clone().unwrap_or_default();
+            // Heal outcome is unconstrained (may hit the tail race and report a
+            // retryable error); the invariant is checked on the end state.
+            drop(heal_res);
+        }
+
+        let final_info = set
+            .get_object_info(bucket, object, &ObjectOptions::default())
+            .await
+            .expect("object must remain readable after the race loop");
+        assert_eq!(
+            final_info.size, PAYLOAD_SIZE as i64,
+            "final current version must be the last committed overwrite"
+        );
+        assert_eq!(
+            final_info.etag.unwrap_or_default(),
+            last_etag,
+            "the racing heal loop must never leave a stale or resurrected current version"
+        );
+    }
 }
@@ -23,6 +23,7 @@
 //! per-version `SetDisks::heal_object`.

 use super::super::*;
+use crate::object_api::ObjectInfo;
 use std::collections::HashSet;
 use std::sync::Mutex;
 use std::sync::atomic::{AtomicBool, AtomicUsize, Ordering};
@@ -39,12 +40,16 @@ const BACKGROUND_WALKDIR_STALL_TIMEOUT: Duration = Duration::from_secs(60);
 /// it must not gate healing logic — the delete-marker vs data path is chosen
 /// inside `ops/heal.rs` from the resolved latest metadata. `version_id` is
 /// normalized (nil/absent UUID => `None`).
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone)]
 pub struct HealWalkVersion {
    /// object key
    pub name: String,
    /// normalized version id (`None` when the version is nil/absent)
    pub version_id: Option<String>,
+    /// version modification time as Unix nanoseconds
+    pub mod_time_unix_nanos: Option<i128>,
+    /// object snapshot for lifecycle evaluation
+    pub lifecycle_object_info: Option<ObjectInfo>,
    /// whether this version is a delete marker (observability only)
    pub is_delete_marker: bool,
 }
@@ -63,6 +68,7 @@ struct HealWalkCollector {
    bucket: String,
    batch_objects: usize,
    version_budget: usize,
+    include_lifecycle_object_info: bool,
    objects: Mutex<Vec<HealWalkObject>>,
    decode_error: Mutex<Option<DiskError>>,
    version_total: AtomicUsize,
@@ -116,10 +122,25 @@ impl HealWalkCollector {

        let mut versions = Vec::with_capacity(fiv.versions.len() + fiv.free_versions.len());
        for fi in fiv.versions.iter().chain(fiv.free_versions.iter()) {
+            let version_uuid = fi.version_id.filter(|version_id| !version_id.is_nil());
+            let lifecycle_object_info = if self.include_lifecycle_object_info {
+                let mut lifecycle_fi = fi.clone();
+                lifecycle_fi.version_id = version_uuid;
+                Some(ObjectInfo::from_file_info(
+                    &lifecycle_fi,
+                    &self.bucket,
+                    &entry.name,
+                    version_uuid.is_some(),
+                ))
+            } else {
+                None
+            };
            versions.push(HealWalkVersion {
                name: entry.name.clone(),
                // Normalize: nil/absent version id => None.
-                version_id: fi.version_id.filter(|u| !u.is_nil()).map(|u| u.to_string()),
+                version_id: version_uuid.map(|u| u.to_string()),
+                mod_time_unix_nanos: fi.mod_time.map(|mod_time| mod_time.unix_timestamp_nanos()),
+                lifecycle_object_info,
                is_delete_marker: fi.deleted,
            });
        }
@@ -173,11 +194,26 @@ impl HealWalkCollector {
                }
            };
            for fi in fiv.versions.iter().chain(fiv.free_versions.iter()) {
-                let vid = fi.version_id.filter(|u| !u.is_nil()).map(|u| u.to_string());
+                let version_uuid = fi.version_id.filter(|version_id| !version_id.is_nil());
+                let vid = version_uuid.map(|u| u.to_string());
                if seen.insert(vid.clone()) {
+                    let lifecycle_object_info = if self.include_lifecycle_object_info {
+                        let mut lifecycle_fi = fi.clone();
+                        lifecycle_fi.version_id = version_uuid;
+                        Some(ObjectInfo::from_file_info(
+                            &lifecycle_fi,
+                            &self.bucket,
+                            &entry.name,
+                            version_uuid.is_some(),
+                        ))
+                    } else {
+                        None
+                    };
                    versions.push(HealWalkVersion {
                        name: entry.name.clone(),
                        version_id: vid,
+                        mod_time_unix_nanos: fi.mod_time.map(|mod_time| mod_time.unix_timestamp_nanos()),
+                        lifecycle_object_info,
                        is_delete_marker: fi.deleted,
                    });
                }
@@ -255,6 +291,7 @@ impl SetDisks {
        forward_to: Option<&str>,
        batch_objects: usize,
        version_budget: usize,
+        include_lifecycle_object_info: bool,
    ) -> disk::error::Result<(Vec<HealWalkVersion>, Option<String>, bool)> {
        assert!(batch_objects >= 2, "heal_walk_versions_page requires batch_objects >= 2");

@@ -264,6 +301,7 @@ impl SetDisks {
            bucket: bucket.to_string(),
            batch_objects,
            version_budget: version_budget.max(1),
+            include_lifecycle_object_info,
            objects: Mutex::new(Vec::new()),
            decode_error: Mutex::new(None),
            version_total: AtomicUsize::new(0),
@@ -347,6 +385,7 @@ mod tests {
            bucket: "bucket".to_string(),
            batch_objects: 2,
            version_budget: 2,
+            include_lifecycle_object_info: false,
            objects: Mutex::new(Vec::new()),
            decode_error: Mutex::new(None),
            version_total: AtomicUsize::new(0),
@@ -388,6 +427,8 @@ mod tests {
        HealWalkVersion {
            name: name.to_string(),
            version_id: Some(id.to_string()),
+            mod_time_unix_nanos: None,
+            lifecycle_object_info: None,
            is_delete_marker: dm,
        }
    }
@@ -491,6 +532,7 @@ mod tests {
            bucket: "bucket".to_string(),
            batch_objects: 1000,
            version_budget: 10_000,
+            include_lifecycle_object_info: false,
            objects: Mutex::new(Vec::new()),
            version_total: AtomicUsize::new(0),
            decode_error: Mutex::new(None),
@@ -567,7 +609,7 @@ mod tests {
        .expect("corrupt test metadata should be written");

        let error = set_disks
-            .heal_walk_versions_page(bucket, "", None, 2, 2)
+            .heal_walk_versions_page(bucket, "", None, 2, 2, false)
            .await
            .expect_err("semantic metadata corruption must fail the heal disk walk");

@@ -66,6 +66,7 @@ impl crate::storage_api_contracts::namespace::NamespaceLocking for SetDisks {
 }

 impl SetDisks {
+    #[allow(dead_code, reason = "lock diagnostics formatter with no caller in this port (backlog#1823)")]
    pub(in crate::set_disk) fn format_lock_error(&self, bucket: &str, object: &str, mode: &str, err: &LockResult) -> String {
        match err {
            LockResult::Timeout => {
@@ -79,6 +80,7 @@ impl SetDisks {
        }
    }

+    #[allow(dead_code, reason = "lock diagnostics formatter with no caller in this port (backlog#1823)")]
    pub(in crate::set_disk) fn format_lock_error_from_error(
        &self,
        bucket: &str,
@@ -143,6 +145,7 @@ impl SetDisks {
        disks
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(in crate::set_disk) async fn get_online_disks(&self) -> Vec<Option<DiskStore>> {
        let snapshot = self.drive_membership_snapshot().await;
        let mut disks = snapshot.strict_online_candidates().into_iter().map(Some).collect::<Vec<_>>();
@@ -153,6 +156,10 @@ impl SetDisks {
        disks
    }

+    #[allow(
+        dead_code,
+        reason = "local-only sibling of the test-covered get_online_disks; no caller in this port (backlog#1823)"
+    )]
    pub(in crate::set_disk) async fn get_online_local_disks(&self) -> Vec<Option<DiskStore>> {
        let snapshot = self.drive_membership_snapshot().await;
        let mut disks = snapshot
@@ -432,6 +439,10 @@ impl SetDisks {
        Ok((disk, fm))
    }

+    #[allow(
+        dead_code,
+        reason = "MinIO-parity healing-disk accessor with no caller in this port (backlog#1823)"
+    )]
    pub(in crate::set_disk) async fn get_online_disk_with_healing(
        &self,
        incl_healing: bool,
@@ -440,6 +451,10 @@ impl SetDisks {
        Ok((new_disks, healing > 0))
    }

+    #[allow(
+        dead_code,
+        reason = "reached only from get_online_disk_with_healing, itself uncalled in this port (backlog#1823)"
+    )]
    pub(in crate::set_disk) async fn get_online_disk_with_healing_and_info(
        &self,
        incl_healing: bool,
@@ -415,6 +415,7 @@ fn reduce_quorum_part_numbers(object_parts: Vec<Vec<String>>, read_quorum: usize
 /// never returned, but flips `is_truncated` to `true` and yields a
 /// `next_upload_id_marker` pointing at the last returned upload so the caller can
 /// resume paging.
+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn paginate_upload_page(remaining: &[MultipartInfo], max_uploads: usize) -> (Vec<MultipartInfo>, bool, Option<String>) {
    let is_truncated = remaining.len() > max_uploads;
    let page: Vec<MultipartInfo> = remaining.iter().take(max_uploads).cloned().collect();
@@ -557,6 +558,7 @@ impl SetDisks {
    }

    #[tracing::instrument(level = "debug", skip(self))]
+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    pub(super) async fn check_upload_id_exists(
        &self,
        bucket: &str,
@@ -1398,7 +1400,7 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for SetDisks {

        let mut count = max_parts;

-        for (i, part) in object_parts.iter().enumerate() {
+        for part in object_parts.iter() {
            if let Some(err) = &part.error {
                warn!("list_object_parts part error: {:?}", &err);
            }
@@ -2041,8 +2043,8 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for SetDisks {
                    && let Err(err) = checksum.add_part(&cs, ext_part.actual_size)
                {
                    error!(
-                        "complete_multipart_upload checksum add_part failed part_id={}, bucket={}, object={}",
-                        p.part_num, bucket, object
+                        "complete_multipart_upload checksum add_part failed part_id={}, bucket={}, object={}, err={}",
+                        p.part_num, bucket, object, err
                    );
                    return Err(Error::InvalidPart(p.part_num, ext_part.etag.clone(), p.etag.clone().unwrap_or_default()));
                }
@@ -2087,8 +2089,8 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for SetDisks {
                }
            } else if let Err(err) = wtcs.matches(&checksum_combined, uploaded_parts.len() as i32) {
                error!(
-                    "complete_multipart_upload checksum matches failed want={}, got={}",
-                    wtcs.encoded, checksum.encoded
+                    "complete_multipart_upload checksum matches failed want={}, got={}, err={}",
+                    wtcs.encoded, checksum.encoded, err
                );
                return Err(Error::other(format!(
                    "complete_multipart_upload checksum matches failed want={}, got={}",
@@ -3507,6 +3507,10 @@ struct TransitionUploadedSaveProbeState {
 }

 #[cfg(test)]
+#[allow(
+    dead_code,
+    reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+)]
 struct TransitionUploadedSaveProbe {
    state: Arc<TransitionUploadedSaveProbeState>,
 }
@@ -3517,6 +3521,10 @@ static TRANSITION_UPLOADED_SAVE_PROBE: std::sync::OnceLock<std::sync::Mutex<Opti

 #[cfg(test)]
 impl TransitionUploadedSaveProbe {
+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn install(bucket: &str, object: &str) -> Self {
        let state = Arc::new(TransitionUploadedSaveProbeState {
            bucket: bucket.to_string(),
@@ -3533,6 +3541,10 @@ impl TransitionUploadedSaveProbe {
        Self { state }
    }

+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn attempts(&self) -> usize {
        self.state.attempts.load(std::sync::atomic::Ordering::Acquire)
    }
@@ -3738,6 +3750,10 @@ struct TransitionCommitBarrierState {
 }

 #[cfg(test)]
+#[allow(
+    dead_code,
+    reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+)]
 struct TransitionCommitBarrier {
    state: Arc<TransitionCommitBarrierState>,
 }
@@ -3748,14 +3764,26 @@ static TRANSITION_COMMIT_BARRIER: std::sync::OnceLock<std::sync::Mutex<Option<Ar

 #[cfg(test)]
 impl TransitionCommitBarrier {
+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn install_before_lock_lost_check(bucket: &str, object: &str) -> Self {
        Self::install_at(bucket, object, TransitionCommitPause::BeforeLockLost)
    }

+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn install(bucket: &str, object: &str) -> Self {
        Self::install_at(bucket, object, TransitionCommitPause::BeforeLeaseValidation)
    }

+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn install_after_lease_check(bucket: &str, object: &str) -> Self {
        Self::install_at(bucket, object, TransitionCommitPause::AfterLeaseValidation)
    }
@@ -3778,12 +3806,20 @@ impl TransitionCommitBarrier {
        Self { state }
    }

+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    async fn wait_until_paused(&self) {
        tokio::time::timeout(Duration::from_secs(30), self.state.arrived.notified())
            .await
            .expect("transition should reach the deterministic commit barrier");
    }

+    #[allow(
+        dead_code,
+        reason = "installed by set_disk tests behind `--features test-util` (backlog#1823)"
+    )]
    fn release(&self) {
        self.state.release.notify_one();
    }
@@ -5620,7 +5656,9 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {
        // TODO: Lifecycle

        let mut version_found = true;
-        let (mut goi, write_quorum, gerr) = self.get_object_info_and_quorum(bucket, object, &opts).await;
+        // delete_object_version below derives its own majority quorum from the
+        // disk array, so the object-derived quorum here is unused.
+        let (mut goi, _write_quorum, gerr) = self.get_object_info_and_quorum(bucket, object, &opts).await;
        if let Some(err) = &gerr
            && goi.name.is_empty()
        {
@@ -5807,6 +5845,14 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {

    #[tracing::instrument(skip(self))]
    async fn add_partial(&self, bucket: &str, object: &str, version_id: &str) -> Result<()> {
+        // MRF journal intent: partial-write recovery must survive a restart
+        // (HS-01); the heal request below remains the in-memory fast path.
+        rustfs_common::mrf_channel::try_send_mrf_intent(
+            rustfs_common::mrf_channel::MrfKind::PartialWrite,
+            bucket,
+            object,
+            uuid::Uuid::try_parse(version_id).ok(),
+        );
        let mut request = rustfs_common::heal_channel::create_heal_request_with_options(
            bucket.to_string(),
            Some(object.to_string()),
@@ -6374,7 +6420,7 @@ impl crate::storage_api_contracts::object::ObjectOperations for SetDisks {
        self.record_capacity_scope_if_needed(opts.capacity_scope_token, &disks);

        for disk in disks.iter() {
-            if let Some(disk) = disk {
+            if disk.is_some() {
                continue;
            }
            let _ = self
@@ -10170,6 +10216,288 @@ mod transition_upload_integrity_tests {
        assert!(backend.contains(remote_object).await, "committed remote object should remain available");
    }

+    /// Compresses `plaintext` with the codec the PUT path uses, so the stored
+    /// bytes round-trip through the read path's decompressor.
+    async fn compress_for_storage(plaintext: &[u8]) -> Vec<u8> {
+        let mut reader = crate::io_support::rio::compression_reader(
+            Cursor::new(plaintext.to_vec()),
+            rustfs_utils::CompressionAlgorithm::default(),
+            false,
+        );
+        let mut compressed = Vec::new();
+        reader.read_to_end(&mut compressed).await.expect("plaintext should compress");
+        assert!(compressed.len() < plaintext.len(), "test payload must actually compress");
+        compressed
+    }
+
+    /// Writes a genuinely compressed object: stored data is `compressed`, and the
+    /// metadata marks it compressed with the plaintext length as its actual size,
+    /// exactly as the app-layer compress path records it.
+    async fn write_compressed_source(
+        set_disks: &Arc<SetDisks>,
+        disk_stores: &[DiskStore],
+        bucket: &str,
+        object: &str,
+        plaintext: &[u8],
+        compressed: &[u8],
+    ) -> ObjectInfo {
+        for disk in disk_stores {
+            disk.make_volume(bucket).await.expect("bucket volume should be created");
+        }
+        let mut user_defined = HashMap::new();
+        rustfs_utils::http::insert_str(
+            &mut user_defined,
+            rustfs_utils::http::SUFFIX_COMPRESSION,
+            crate::io_support::rio::compression_metadata_value(rustfs_utils::CompressionAlgorithm::default()),
+        );
+        rustfs_utils::http::insert_str(&mut user_defined, rustfs_utils::http::SUFFIX_ACTUAL_SIZE, plaintext.len().to_string());
+        let stream = crate::io_support::rio::HashReader::from_stream(
+            Cursor::new(compressed.to_vec()),
+            compressed.len() as i64,
+            plaintext.len() as i64,
+            None,
+            None,
+            false,
+        )
+        .expect("hash reader over compressed bytes");
+        let mut reader = PutObjReader::new(stream);
+        set_disks
+            .put_object(
+                bucket,
+                object,
+                &mut reader,
+                &ObjectOptions {
+                    no_lock: true,
+                    user_defined,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("compressed object should be written")
+    }
+
+    async fn read_transitioned(
+        set_disks: &Arc<SetDisks>,
+        bucket: &str,
+        object: &str,
+        range: Option<HTTPRangeSpec>,
+        opts: &ObjectOptions,
+    ) -> (Vec<u8>, i64) {
+        let mut reader = set_disks
+            .get_object_reader(bucket, object, range, HeaderMap::new(), opts)
+            .await
+            .expect("transitioned object reader should open");
+        let published_size = reader.object_info.size;
+        let mut body = Vec::new();
+        reader
+            .stream
+            .read_to_end(&mut body)
+            .await
+            .expect("transitioned body should drain");
+        (body, published_size)
+    }
+
+    /// Transition uploads the object's STORED bytes, so a tiered read has to
+    /// apply the same transform an erasure read would. #6107 routed this path
+    /// through `ReadPlan` to stop serving an encrypted object's ciphertext;
+    /// compression rides the same plan, and nothing pinned it (backlog#1851).
+    /// Without the transform this GET returns the compressed bytes under the
+    /// compressed size — silent corruption for every client of a compressed
+    /// object that ILM has moved to a warm tier.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn transitioned_compressed_object_get_returns_plaintext() {
+        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
+        let bucket = "transitioned-compressed-get-bucket";
+        let object = "object.txt";
+        let plaintext = b"transitioned compressed objects must decompress on read ".repeat(20_000);
+        let compressed = compress_for_storage(&plaintext).await;
+        let original = write_compressed_source(&set_disks, &disk_stores, bucket, object, &plaintext, &compressed).await;
+
+        let opts = ObjectOptions {
+            no_lock: true,
+            ..Default::default()
+        };
+        let (local_body, local_size) = read_transitioned(&set_disks, bucket, object, None, &opts).await;
+        assert_eq!(local_body, plaintext, "control: the pre-transition read must decompress");
+        assert_eq!(
+            local_size,
+            plaintext.len() as i64,
+            "control: the pre-transition read publishes the plaintext size"
+        );
+
+        let tier_name = format!("COLDTIER{}", &Uuid::new_v4().simple().to_string()[..8]).to_uppercase();
+        let backend = register_mock_tier(&runtime_sources::global_tier_config_mgr(), &tier_name).await;
+        set_disks
+            .transition_object(bucket, object, &transition_options(&original, tier_name))
+            .await
+            .expect("transition should commit");
+
+        let put_versions = backend.put_versions().await;
+        assert_eq!(put_versions.len(), 1, "transition should upload one remote candidate");
+        let remote_bytes = backend
+            .bytes(&put_versions[0].0)
+            .await
+            .expect("remote candidate should be stored");
+        assert_eq!(
+            remote_bytes, compressed,
+            "transition uploads the stored representation; the read side is what has to decode it"
+        );
+
+        let (body, published_size) = read_transitioned(&set_disks, bucket, object, None, &opts).await;
+        assert_eq!(body, plaintext, "a tiered read must return the object's content, not its stored bytes");
+        assert_eq!(
+            published_size,
+            plaintext.len() as i64,
+            "a tiered read must publish the plaintext size, not the compressed one"
+        );
+    }
+
+    /// A ranged tiered read is expressed in plaintext coordinates, so the plan
+    /// has to translate it into the remote copy's compressed extent and skip
+    /// into the decompressed stream — the same translation the erasure path does.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn transitioned_compressed_object_range_get_returns_plaintext_slice() {
+        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
+        let bucket = "transitioned-compressed-range-bucket";
+        let object = "object.txt";
+        let plaintext = b"ranged reads of transitioned compressed objects must land in plaintext ".repeat(20_000);
+        let compressed = compress_for_storage(&plaintext).await;
+        let original = write_compressed_source(&set_disks, &disk_stores, bucket, object, &plaintext, &compressed).await;
+
+        let tier_name = format!("COLDTIER{}", &Uuid::new_v4().simple().to_string()[..8]).to_uppercase();
+        register_mock_tier(&runtime_sources::global_tier_config_mgr(), &tier_name).await;
+        set_disks
+            .transition_object(bucket, object, &transition_options(&original, tier_name))
+            .await
+            .expect("transition should commit");
+
+        let opts = ObjectOptions {
+            no_lock: true,
+            ..Default::default()
+        };
+        // Deliberately past the compressed size, so a range still measured in
+        // stored coordinates could not produce this slice.
+        let start = compressed.len() as i64 + 4096;
+        let end = start + 511;
+        let range = HTTPRangeSpec {
+            is_suffix_length: false,
+            start,
+            end,
+        };
+        let (body, published_size) = read_transitioned(&set_disks, bucket, object, Some(range), &opts).await;
+
+        let expected = &plaintext[start as usize..=end as usize];
+        assert_eq!(body, expected, "a ranged tiered read must return that plaintext slice");
+        assert_eq!(published_size, expected.len() as i64, "a ranged tiered read publishes the slice length");
+    }
+
+    /// The restore copy-back re-writes the object under its original metadata,
+    /// which still says "compressed". It therefore has to keep receiving the
+    /// STORED bytes: `restore_request_active` holds it on the plan's `Plain`
+    /// branch, and decompressing there would write plaintext under compressed
+    /// metadata.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn restore_read_of_transitioned_compressed_object_keeps_stored_bytes() {
+        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
+        let bucket = "transitioned-compressed-restore-bucket";
+        let object = "object.txt";
+        let plaintext = b"restore copy-back must keep the stored representation intact ".repeat(20_000);
+        let compressed = compress_for_storage(&plaintext).await;
+        let original = write_compressed_source(&set_disks, &disk_stores, bucket, object, &plaintext, &compressed).await;
+
+        let tier_name = format!("COLDTIER{}", &Uuid::new_v4().simple().to_string()[..8]).to_uppercase();
+        register_mock_tier(&runtime_sources::global_tier_config_mgr(), &tier_name).await;
+        set_disks
+            .transition_object(bucket, object, &transition_options(&original, tier_name))
+            .await
+            .expect("transition should commit");
+
+        let oi = set_disks
+            .get_object_info(
+                bucket,
+                object,
+                &ObjectOptions {
+                    no_lock: true,
+                    ..Default::default()
+                },
+            )
+            .await
+            .expect("transitioned metadata should resolve");
+        let restore_opts = ObjectOptions {
+            no_lock: true,
+            part_number: Some(1),
+            transition: TransitionOptions {
+                restore_request: s3s::dto::RestoreRequest {
+                    days: Some(1),
+                    ..Default::default()
+                },
+                ..Default::default()
+            },
+            ..Default::default()
+        };
+        let mut reader = get_transitioned_object_reader_with_tier_manager(
+            bucket,
+            object,
+            &None,
+            &HeaderMap::new(),
+            &oi,
+            &restore_opts,
+            &set_disks.ctx.tier_config_mgr(),
+            set_disks.ctx.object_encryption_resolver(),
+        )
+        .await
+        .expect("restore read of the tiered copy should open");
+        let published_size = reader.object_info.size;
+        let mut body = Vec::new();
+        reader.stream.read_to_end(&mut body).await.expect("restore body should drain");
+
+        assert_eq!(body, compressed, "a restore read must copy the stored bytes back verbatim");
+        assert_eq!(
+            published_size,
+            compressed.len() as i64,
+            "a restore read must keep publishing the stored size"
+        );
+    }
+
+    /// Plain objects must keep streaming the remote bytes through untouched:
+    /// their plan is `Plain`, so the tiered read stays byte-identical.
+    #[tokio::test]
+    #[serial_test::serial]
+    async fn transitioned_plain_object_get_is_unchanged() {
+        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
+        let bucket = "transitioned-plain-get-bucket";
+        let object = "object.bin";
+        let payload = b"plain transitioned objects must keep reading back byte-identical ".repeat(1024);
+        let original = write_source(&set_disks, &disk_stores, bucket, object, &payload).await;
+
+        let tier_name = format!("COLDTIER{}", &Uuid::new_v4().simple().to_string()[..8]).to_uppercase();
+        register_mock_tier(&runtime_sources::global_tier_config_mgr(), &tier_name).await;
+        set_disks
+            .transition_object(bucket, object, &transition_options(&original, tier_name))
+            .await
+            .expect("transition should commit");
+
+        let opts = ObjectOptions {
+            no_lock: true,
+            ..Default::default()
+        };
+        let (body, published_size) = read_transitioned(&set_disks, bucket, object, None, &opts).await;
+        assert_eq!(body, payload);
+        assert_eq!(published_size, payload.len() as i64);
+
+        let range = HTTPRangeSpec {
+            is_suffix_length: false,
+            start: 100,
+            end: 611,
+        };
+        let (ranged_body, ranged_size) = read_transitioned(&set_disks, bucket, object, Some(range), &opts).await;
+        assert_eq!(ranged_body, &payload[100..=611]);
+        assert_eq!(ranged_size, payload.len() as i64, "a plain ranged read keeps publishing the object size");
+    }
+
    async fn corrupt_beyond_read_quorum(
        temp_dirs: &[tempfile::TempDir],
        bucket: &str,
@@ -116,6 +116,7 @@ impl SetDisks {
            .then_some(GET_METADATA_CACHE_REASON_DIST_ERASURE)
    }

+    #[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
    async fn cached_get_object_fileinfo(&self, bucket: &str, object: &str) -> Option<Arc<GetObjectMetadataCacheEntry>> {
        match self.lookup_cached_get_object_fileinfo(bucket, object).await {
            MetadataCacheLookup::Hit(entry) => Some(entry),
@@ -1076,6 +1077,15 @@ impl SetDisks {
                            "Recoverable decode error triggered read repair"
                        );
                        let version_id = fi.version_id.as_ref().map(ToString::to_string);
+                        // MRF journal intent: keeps a durable Urgent ECDecode
+                        // request alive across restarts even when the in-memory
+                        // read-repair request is dropped or lost (HS-01).
+                        rustfs_common::mrf_channel::try_send_mrf_intent(
+                            rustfs_common::mrf_channel::MrfKind::DecodeFailure,
+                            bucket,
+                            object,
+                            fi.version_id,
+                        );
                        submit_read_repair_heal(
                            bucket,
                            object,
@@ -1826,6 +1836,7 @@ fn get_object_metadata_cache_request_bypass_reason(bucket: &str, opts: &ObjectOp
        .then_some(GET_METADATA_CACHE_REASON_META_BUCKET)
 }

+#[allow(dead_code, reason = "asserted by this file's tests (backlog#1823)")]
 fn is_get_object_metadata_cache_request_eligible(bucket: &str, opts: &ObjectOptions, read_data: bool) -> bool {
    get_object_metadata_cache_request_bypass_reason(bucket, opts, read_data).is_none()
 }
@@ -3935,7 +3946,7 @@ mod tests {
    }

    #[test]
-    fn metadata_early_stop_bounded_fanout_defaults_to_disabled() {
+    fn metadata_early_stop_bounded_fanout_defaults_to_enabled() {
        temp_env::with_vars(
            [
                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_ENABLE, Some("true")),
@@ -3944,20 +3955,20 @@ mod tests {
            ],
            || {
                assert!(is_get_metadata_data_read_early_stop_enabled());
-                assert!(!is_get_metadata_early_stop_bounded_fanout_enabled());
+                assert!(is_get_metadata_early_stop_bounded_fanout_enabled());
            },
        );
-        temp_env::with_vars([(ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT, Some("true"))], || {
-            assert!(is_get_metadata_early_stop_bounded_fanout_enabled());
+        temp_env::with_vars([(ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT, Some("false"))], || {
+            assert!(!is_get_metadata_early_stop_bounded_fanout_enabled());
        });
        temp_env::with_vars(
            [
                (ENV_RUSTFS_GET_METADATA_DATA_READ_EARLY_STOP_ENABLE, Some("false")),
-                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT, Some("false")),
+                (ENV_RUSTFS_GET_METADATA_EARLY_STOP_BOUNDED_FANOUT, Some("true")),
            ],
            || {
                assert!(!is_get_metadata_data_read_early_stop_enabled());
-                assert!(!is_get_metadata_early_stop_bounded_fanout_enabled());
+                assert!(is_get_metadata_early_stop_bounded_fanout_enabled());
            },
        );
    }
@@ -18,6 +18,7 @@ use tracing::trace;

 const LOG_COMPONENT_ECSTORE: &str = "ecstore";
 const LOG_SUBSYSTEM_HEAL: &str = "heal";
+const EVENT_HEAL_ABANDONED_PARTS: &str = "heal_abandoned_parts";
 const EVENT_HEAL_FORMAT_COMPLETED: &str = "heal_format_completed";
 const EVENT_HEAL_OBJECT_STARTED: &str = "heal_object_started";

@@ -256,13 +257,40 @@ impl ECStore {

    #[instrument(skip(self))]
    pub(super) async fn handle_check_abandoned_parts(&self, bucket: &str, object: &str, opts: &HealOpts) -> Result<()> {
-        let _ = (bucket, object, opts);
-        // Stale multipart reconciliation is already owned by the lifecycle-driven
-        // background cleanup path in `bucket_lifecycle_ops.rs`. There is currently
-        // no stable object-heal contract that should fan this request out through
-        // pool/set storage layers, so keep the placeholder explicit at the ECStore
-        // boundary instead of dispatching into lower layers.
-        Err(StorageError::NotImplemented)
+        let object = encode_dir_object(object);
+        let pools = self.get_pools_for_heal_object(opts)?;
+
+        let mut futures = Vec::with_capacity(pools.len());
+        for pool in pools.iter() {
+            futures.push(pool.check_abandoned_parts(bucket, &object, opts));
+        }
+
+        let mut first_error = None;
+        for result in join_all(futures).await {
+            if let Err(err) = result
+                && first_error.is_none()
+            {
+                first_error = Some(err);
+            }
+        }
+
+        if let Some(err) = first_error {
+            return Err(err);
+        }
+
+        trace!(
+            event = EVENT_HEAL_ABANDONED_PARTS,
+            component = LOG_COMPONENT_ECSTORE,
+            subsystem = LOG_SUBSYSTEM_HEAL,
+            state = "completed",
+            result = "ok",
+            bucket,
+            object,
+            dry_run = opts.dry_run,
+            "Heal abandoned parts completed"
+        );
+
+        Ok(())
    }
 }

@@ -34,6 +34,7 @@ impl ECStore {
        forward_to: Option<&str>,
        batch_objects: usize,
        version_budget: usize,
+        include_lifecycle_object_info: bool,
    ) -> Result<(Vec<HealWalkVersion>, Option<String>, bool)> {
        if pool_idx >= self.pools.len() || set_idx >= self.pools[pool_idx].disk_set.len() {
            return Err(Error::other(format!(
@@ -43,7 +44,7 @@ impl ECStore {
        }

        self.pools[pool_idx].disk_set[set_idx]
-            .heal_walk_versions_page(bucket, prefix, forward_to, batch_objects, version_budget)
+            .heal_walk_versions_page(bucket, prefix, forward_to, batch_objects, version_budget, include_lifecycle_object_info)
            .await
            .map_err(Error::from)
    }
@@ -216,6 +216,16 @@ impl std::fmt::Debug for ECStore {
 /// These delegate to the process-global statics. No local state — the globals
 /// remain the single source of truth until the migration is complete.
 impl ECStore {
+    /// Every erasure set across all pools, pool-major order.
+    ///
+    /// Read-only queries that must consult each set's own copy of a
+    /// per-bucket object (e.g. the scanner's `.usage-cache.bin`) iterate
+    /// this instead of the hash-routed store path, which would always land
+    /// on one set (rustfs/backlog#1872).
+    pub fn all_set_disks(&self) -> Vec<Arc<crate::set_disk::SetDisks>> {
+        self.pools.iter().flat_map(|pool| pool.disk_set.iter().cloned()).collect()
+    }
+
    /// Get server configuration (delegates to global)
    pub fn get_server_config(&self) -> Option<Config> {
        runtime_sources::server_config()
@@ -89,6 +89,8 @@ async-trait = { workspace = true }
 futures = { workspace = true }
 metrics = { workspace = true }
 base64 = { workspace = true }
+bytes = { workspace = true }
+crc-fast = { workspace = true }

 [dev-dependencies]
 serde_json = { workspace = true, features = ["raw_value"] }
@@ -66,21 +66,37 @@ struct HealTaskStatusPayload<'a> {
    summary: &'a str,
    items: &'a [HealResultItem],
    truncated: bool,
+    /// Cursor for incremental consumption (HS-06): sequence of the next item
+    /// to be produced. Absent on responses without sequencing (0).
+    #[serde(skip_serializing_if = "u64_is_zero")]
+    next_seq: u64,
+    /// Oldest sequence still retained; with `truncated`, tells a lagging
+    /// client where to restart its cursor.
+    #[serde(skip_serializing_if = "u64_is_zero")]
+    min_seq: u64,
    #[serde(skip_serializing_if = "Option::is_none")]
    progress: Option<&'a HealProgress>,
 }

+fn u64_is_zero(value: &u64) -> bool {
+    *value == 0
+}
+
 fn encode_heal_task_status_payload(
    summary: &str,
    mut items: Vec<HealResultItem>,
    progress: Option<&HealProgress>,
    mut truncated: bool,
+    next_seq: u64,
+    min_seq: u64,
 ) -> Result<(Vec<u8>, bool)> {
    loop {
        let data = serde_json::to_vec(&HealTaskStatusPayload {
            summary,
            items: &items,
            truncated,
+            next_seq,
+            min_seq,
            progress,
        })
        .map_err(|e| Error::Serialization(format!("failed to serialize heal task status: {e}")))?;
@@ -109,8 +125,10 @@ fn encode_heal_status_response(
    progress: Option<&HealProgress>,
    detail: Option<String>,
    truncated: bool,
+    next_seq: u64,
+    min_seq: u64,
 ) -> Result<(Vec<u8>, Option<String>)> {
-    let (data, truncated) = encode_heal_task_status_payload(summary, items, progress, truncated)?;
+    let (data, truncated) = encode_heal_task_status_payload(summary, items, progress, truncated, next_seq, min_seq)?;
    Ok((data, heal_status_detail(detail, truncated)))
 }

@@ -138,8 +156,19 @@ impl HealChannelProcessor {

    /// Execute a token query directly against the manager.
    pub async fn execute_query_request(&self, heal_path: String, client_token: String) -> Result<HealChannelResponse> {
+        self.execute_query_request_since(heal_path, client_token, None).await
+    }
+
+    /// Incremental variant of [`Self::execute_query_request`] (HS-06).
+    pub async fn execute_query_request_since(
+        &self,
+        heal_path: String,
+        client_token: String,
+        since_seq: Option<u64>,
+    ) -> Result<HealChannelResponse> {
        let (response_tx, response_rx) = oneshot::channel();
-        self.process_query_request(heal_path, client_token, response_tx).await?;
+        self.process_query_request(heal_path, client_token, since_seq, response_tx)
+            .await?;
        response_rx
            .await
            .map_err(|err| Error::other(format!("heal query channel closed: {err}")))?
@@ -262,8 +291,12 @@ impl HealChannelProcessor {
            HealChannelCommand::Query {
                heal_path,
                client_token,
+                since_seq,
                response_tx,
-            } => self.process_query_request(heal_path, client_token, response_tx).await,
+            } => {
+                self.process_query_request(heal_path, client_token, since_seq, response_tx)
+                    .await
+            }
            HealChannelCommand::Cancel {
                heal_path,
                client_token,
@@ -384,6 +417,7 @@ impl HealChannelProcessor {
        &self,
        heal_path: String,
        client_token: String,
+        since_seq: Option<u64>,
        response_tx: oneshot::Sender<std::result::Result<HealChannelResponse, String>>,
    ) -> Result<()> {
        debug!(
@@ -398,72 +432,118 @@ impl HealChannelProcessor {
        );

        let report = if heal_path.trim_matches('/').is_empty() {
-            self.heal_manager.get_task_report(&client_token).await
+            self.heal_manager.get_task_report_since(&client_token, since_seq).await
        } else {
-            self.heal_manager.get_task_report_for_path(&heal_path, &client_token).await
+            self.heal_manager
+                .get_task_report_for_path_since(&heal_path, &client_token, since_seq)
+                .await
        };

-        let (summary, detail, items, truncated, progress) = match report {
+        let (summary, detail, items, truncated, progress, next_seq, min_seq) = match report {
            Ok(HealTaskReport {
                status: HealTaskStatus::Pending | HealTaskStatus::Running,
                result_items,
                result_items_truncated,
                progress,
-            }) => ("running".to_string(), None, result_items, result_items_truncated, progress),
+                next_seq,
+                min_seq,
+            }) => (
+                "running".to_string(),
+                None,
+                result_items,
+                result_items_truncated,
+                progress,
+                next_seq,
+                min_seq,
+            ),
            Ok(HealTaskReport {
                status: HealTaskStatus::Retrying { error, retry_attempt },
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            }) => (
                "running".to_string(),
                Some(format!("heal task retrying after recoverable failure, attempt {retry_attempt}: {error}")),
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            ),
            Ok(HealTaskReport {
                status: HealTaskStatus::Completed,
                result_items,
                result_items_truncated,
                progress,
-            }) => ("finished".to_string(), None, result_items, result_items_truncated, progress),
+                next_seq,
+                min_seq,
+            }) => (
+                "finished".to_string(),
+                None,
+                result_items,
+                result_items_truncated,
+                progress,
+                next_seq,
+                min_seq,
+            ),
            Ok(HealTaskReport {
                status: HealTaskStatus::Cancelled,
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            }) => (
                "stopped".to_string(),
                Some("heal task cancelled".to_string()),
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            ),
            Ok(HealTaskReport {
                status: HealTaskStatus::Timeout,
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            }) => (
                "stopped".to_string(),
                Some("heal task timed out".to_string()),
                result_items,
                result_items_truncated,
                progress,
+                next_seq,
+                min_seq,
            ),
            Ok(HealTaskReport {
                status: HealTaskStatus::Failed { error },
                result_items,
                result_items_truncated,
                progress,
-            }) => ("stopped".to_string(), Some(error), result_items, result_items_truncated, progress),
+                next_seq,
+                min_seq,
+            }) => (
+                "stopped".to_string(),
+                Some(error),
+                result_items,
+                result_items_truncated,
+                progress,
+                next_seq,
+                min_seq,
+            ),
            Err(crate::Error::TaskNotFound { .. }) => (
                "notFound".to_string(),
                Some("heal task not found or expired".to_string()),
                Vec::new(),
                false,
                None,
+                0,
+                0,
            ),
            Err(crate::Error::InvalidClientToken) => {
                let response = HealChannelResponse {
@@ -490,7 +570,8 @@ impl HealChannelProcessor {
            }
        };

-        let (data, detail) = encode_heal_status_response(&summary, items, progress.as_ref(), detail, truncated)?;
+        let (data, detail) =
+            encode_heal_status_response(&summary, items, progress.as_ref(), detail, truncated, next_seq, min_seq)?;

        let response = HealChannelResponse {
            request_id: client_token,
@@ -612,7 +693,8 @@ impl HealChannelProcessor {
            HealRequestSource::Admin
            | HealRequestSource::AutoHeal
            | HealRequestSource::Internal
-            | HealRequestSource::ReadRepair => true,
+            | HealRequestSource::ReadRepair
+            | HealRequestSource::Mrf => true,
        });

        // Build HealOptions with all available fields
@@ -767,6 +849,7 @@ mod tests {
            _bucket: &str,
            _prefix: &str,
            _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
        ) -> crate::Result<(Vec<crate::heal::storage::HealListItem>, Option<String>, bool)> {
            Ok((vec![], None, false))
        }
@@ -803,7 +886,7 @@ mod tests {
            ..Default::default()
        }];

-        let (data, detail) = encode_heal_status_response("running", items, None, None, false).unwrap();
+        let (data, detail) = encode_heal_status_response("running", items, None, None, false, 0, 0).unwrap();

        assert!(data.len() <= MAX_HEAL_STATUS_PAYLOAD_SIZE);
        let payload: serde_json::Value = serde_json::from_slice(&data).unwrap();
@@ -1573,7 +1656,7 @@ mod tests {
        let (tx, rx) = oneshot::channel();

        processor
-            .process_query_request("bucket".to_string(), "completed-token".to_string(), tx)
+            .process_query_request("bucket".to_string(), "completed-token".to_string(), None, tx)
            .await
            .expect("query should process");

@@ -1608,7 +1691,7 @@ mod tests {
        let (tx, rx) = oneshot::channel();

        processor
-            .process_query_request("bucket".to_string(), task_id.clone(), tx)
+            .process_query_request("bucket".to_string(), task_id.clone(), None, tx)
            .await
            .expect("query should process");

@@ -1641,7 +1724,7 @@ mod tests {
        let (tx, rx) = oneshot::channel();

        processor
-            .process_query_request("bucket".to_string(), "wrong-token".to_string(), tx)
+            .process_query_request("bucket".to_string(), "wrong-token".to_string(), None, tx)
            .await
            .expect("query should process");

@@ -1666,7 +1749,7 @@ mod tests {
        let (tx, rx) = oneshot::channel();

        processor
-            .process_query_request(String::new(), "wrong-token".to_string(), tx)
+            .process_query_request(String::new(), "wrong-token".to_string(), None, tx)
            .await
            .expect("query should process");

@@ -1703,7 +1786,7 @@ mod tests {
        let (tx, rx) = oneshot::channel();

        processor
-            .process_query_request(String::new(), task_id.clone(), tx)
+            .process_query_request(String::new(), task_id.clone(), None, tx)
            .await
            .expect("query should process");

@@ -23,13 +23,14 @@ use crate::heal::{
 };
 use crate::{Error, Result};
 use futures::{StreamExt, stream::FuturesUnordered};
-use metrics::gauge;
+use metrics::{counter, gauge};
 use rustfs_common::heal_channel::{HealOpts, HealRequestSource, HealScanMode};
 use rustfs_madmin::heal_commands::HealResultItem;
 use std::sync::{
    Arc,
    atomic::{AtomicUsize, Ordering},
 };
+use std::time::{Duration, UNIX_EPOCH};
 use tokio::sync::{RwLock, Semaphore};
 use tracing::{debug, error, warn};

@@ -47,6 +48,21 @@ enum HealObjectOutcome {
    Failed,
 }

+fn result_object_size_u64(result: &HealResultItem) -> u64 {
+    u64::try_from(result.object_size).unwrap_or(u64::MAX)
+}
+
+const NEW_VERSION_SKIP_GRACE_SECS: u64 = 60;
+const NANOS_PER_SECOND: i128 = 1_000_000_000;
+
+fn should_skip_new_version(mod_time_unix_nanos: Option<i128>, started_at_secs: u64) -> bool {
+    let Some(mod_time_unix_nanos) = mod_time_unix_nanos else {
+        return false;
+    };
+    let cutoff_secs = started_at_secs.saturating_add(NEW_VERSION_SKIP_GRACE_SECS);
+    mod_time_unix_nanos > i128::from(cutoff_secs).saturating_mul(NANOS_PER_SECOND)
+}
+
 struct PageConcurrencyGuard {
    in_flight: Arc<AtomicUsize>,
    set_label: String,
@@ -492,6 +508,7 @@ impl ErasureSetHealer {
                    &mut skipped_objects,
                    resume_manager,
                    checkpoint_manager,
+                    state.start_time,
                )
                .await;

@@ -658,6 +675,7 @@ impl ErasureSetHealer {
        skipped_objects: &mut u64,
        resume_manager: &ResumeManager,
        checkpoint_manager: &CheckpointManager,
+        started_at_secs: u64,
    ) -> Result<()> {
        debug!(
            target: "rustfs::heal::erasure_healer",
@@ -710,6 +728,7 @@ impl ErasureSetHealer {
        // The end-of-pass summary reports the full failed/skipped counts.
        let mut transient_skip_samples_logged = 0_u64;
        let mut failure_samples_logged = 0_u64;
+        let mut bytes_processed = self.progress.read().await.bytes_processed;

        // backlog#920: select the per-erasure-set DISK-WALK union enumerator when
        // the scan is Deep OR the request came from AutoHeal — these are the paths
@@ -718,17 +737,25 @@ impl ErasureSetHealer {
        // which stays the default.
        let use_disk_walk =
            matches!(self.heal_opts.scan_mode, HealScanMode::Deep) || matches!(self.source, HealRequestSource::AutoHeal);
+        let lifecycle_expiry_context = self.storage.load_heal_lifecycle_expiry_context(bucket).await?;
+        let include_lifecycle_object_info = lifecycle_expiry_context.is_some();

        loop {
            self.verify_replacement_identity_fence("page scan").await?;
            // Get one page of object versions
            let (objects, next_token, is_truncated) = if use_disk_walk {
                self.storage
-                    .list_versions_for_heal_page_disk_walk(set_disk_id, bucket, "", continuation_token.as_deref())
+                    .list_versions_for_heal_page_disk_walk(
+                        set_disk_id,
+                        bucket,
+                        "",
+                        continuation_token.as_deref(),
+                        include_lifecycle_object_info,
+                    )
                    .await?
            } else {
                self.storage
-                    .list_objects_for_heal_page(bucket, "", continuation_token.as_deref())
+                    .list_objects_for_heal_page(bucket, "", continuation_token.as_deref(), include_lifecycle_object_info)
                    .await?
            };
            let page_is_empty = objects.is_empty();
@@ -736,6 +763,7 @@ impl ErasureSetHealer {
            let page_resume_index = *current_object_index;
            let semaphore = Arc::new(Semaphore::new(page_concurrency_limit));
            let mut page_tasks = FuturesUnordered::new();
+            let mut completed_in_page = 0usize;

            // Capture the last version identity of this page for the anti-loop guard.
            let page_last = objects.last().map(|item| (item.name.clone(), item.version_id.clone()));
@@ -751,6 +779,75 @@ impl ErasureSetHealer {
                    continue;
                }

+                if should_skip_new_version(item.mod_time_unix_nanos, started_at_secs) {
+                    checkpoint_manager.add_processed_object(key).await?;
+                    *processed_objects = processed_objects.saturating_add(1);
+                    completed_in_page = completed_in_page.saturating_add(1);
+                    counter!("rustfs_heal_skipped_new_versions_total").increment(1);
+                    {
+                        let mut progress = self.progress.write().await;
+                        progress.record_skipped_new_version();
+                        progress.set_current_object(Some(format!("skipped_new: {bucket}/{}", item.name)));
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                    }
+                    debug!(
+                        target: "rustfs::heal::erasure_healer",
+                        event = EVENT_HEAL_ERASURE_OBJECT_STATE,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
+                        set_disk_id,
+                        bucket,
+                        object = %item.name,
+                        version_id = ?item.version_id,
+                        state = "skipped_new_version",
+                        "Erasure set object version skipped because it was written after heal started"
+                    );
+                    if completed_in_page.is_multiple_of(100) {
+                        checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
+                    }
+                    continue;
+                }
+
+                if let Some(context) = lifecycle_expiry_context.as_ref()
+                    && self
+                        .storage
+                        .enqueue_heal_lifecycle_expiry(
+                            context,
+                            bucket,
+                            &item.name,
+                            item.version_id.as_deref(),
+                            item.lifecycle_object_info.as_ref(),
+                        )
+                        .await?
+                {
+                    checkpoint_manager.add_processed_object(key).await?;
+                    *processed_objects = processed_objects.saturating_add(1);
+                    completed_in_page = completed_in_page.saturating_add(1);
+                    counter!("rustfs_heal_skipped_ilm_expired_total").increment(1);
+                    {
+                        let mut progress = self.progress.write().await;
+                        progress.record_skipped_ilm_expired();
+                        progress.set_current_object(Some(format!("skipped_ilm: {bucket}/{}", item.name)));
+                        progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                    }
+                    debug!(
+                        target: "rustfs::heal::erasure_healer",
+                        event = EVENT_HEAL_ERASURE_OBJECT_STATE,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_ERASURE_HEALER,
+                        set_disk_id,
+                        bucket,
+                        object = %item.name,
+                        version_id = ?item.version_id,
+                        state = "skipped_ilm_expired",
+                        "Erasure set object version skipped because lifecycle expiry was queued"
+                    );
+                    if completed_in_page.is_multiple_of(100) {
+                        checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
+                    }
+                    continue;
+                }
+
                resume_manager
                    .set_current_item(Some(bucket.to_string()), Some(item.name.clone()))
                    .await?;
@@ -777,7 +874,7 @@ impl ErasureSetHealer {

                    let _permit = match permit {
                        Ok(permit) => permit,
-                        Err(err) => return (dedup_key, object_name, version_id, Err(err)),
+                        Err(err) => return (dedup_key, object_name, version_id, (0, Err(err))),
                    };

                    let _in_flight_guard = PageConcurrencyGuard::new(in_flight, set_label);
@@ -788,7 +885,7 @@ impl ErasureSetHealer {
                    // recorded as skipped-ok rather than failed. The delete-marker
                    // vs data path is chosen internally in ops/heal.rs.
                    let result = if cancel_token.is_cancelled() {
-                        Err(Error::TaskCancelled)
+                        (0, Err(Error::TaskCancelled))
                    } else {
                        match storage
                            .heal_object(&bucket_name, &object_name, version_id.as_deref(), &heal_opts)
@@ -797,8 +894,9 @@ impl ErasureSetHealer {
                            Ok((result, None))
                                if target_outcomes_complete(&result, &target_endpoints) =>
                            {
+                                let object_size = result_object_size_u64(&result);
                                if !replacement_commit_evidence_required {
-                                    Ok(true)
+                                    (object_size, Ok(true))
                                } else {
                                    match storage
                                        .replacement_targets_have_version(
@@ -810,27 +908,42 @@ impl ErasureSetHealer {
                                        )
                                        .await
                                    {
-                                        Ok(true) => Ok(true),
-                                        Ok(false) => Err(Error::transient_skip(format!(
+                                        Ok(true) => (object_size, Ok(true)),
+                                        Ok(false) => (object_size, Err(Error::transient_skip(format!(
                                            "Skipped heal for {bucket_name}/{object_name} because replacement target readback did not confirm the committed version"
-                                        ))),
-                                        Err(err) => Err(Error::transient_skip(format!(
+                                        )))),
+                                        Err(err) => (object_size, Err(Error::transient_skip(format!(
                                            "Skipped heal for {bucket_name}/{object_name} because replacement target readback failed: {err}"
-                                        ))),
+                                        )))),
                                    }
                                }
-                            }
-                            Ok((_result, None)) if !target_endpoints.is_empty() => Err(Error::transient_skip(format!(
-                                "Skipped heal for {bucket_name}/{object_name} because a replacement target was not committed"
-                            ))),
-                            Ok((_result, None)) => Ok(true),
-                            Ok((_, Some(err))) if is_missing_object_dir_heal_result(&object_name, &err) => Ok(false),
-                            Ok((_, Some(err))) | Err(err) => match Self::classify_heal_object_error(&err) {
-                                HealObjectOutcome::Absent => Ok(false),
-                                HealObjectOutcome::Transient => Err(Error::transient_skip(format!(
-                                    "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
+                            },
+                            Ok((result, None)) if !target_endpoints.is_empty() => (
+                                result_object_size_u64(&result),
+                                Err(Error::transient_skip(format!(
+                                    "Skipped heal for {bucket_name}/{object_name} because a replacement target was not committed"
                                ))),
-                                HealObjectOutcome::Failed => Err(err),
+                            ),
+                            Ok((result, None)) => (result_object_size_u64(&result), Ok(true)),
+                            Ok((result, Some(err))) if is_missing_object_dir_heal_result(&object_name, &err) => {
+                                (result_object_size_u64(&result), Ok(false))
+                            }
+                            Ok((result, Some(err))) => {
+                                let object_size = result_object_size_u64(&result);
+                                match Self::classify_heal_object_error(&err) {
+                                    HealObjectOutcome::Absent => (object_size, Ok(false)),
+                                    HealObjectOutcome::Transient => (object_size, Err(Error::transient_skip(format!(
+                                        "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
+                                    )))),
+                                    HealObjectOutcome::Failed => (object_size, Err(err)),
+                                }
+                            }
+                            Err(err) => match Self::classify_heal_object_error(&err) {
+                                HealObjectOutcome::Absent => (0, Ok(false)),
+                                HealObjectOutcome::Transient => (0, Err(Error::transient_skip(format!(
+                                    "Skipped heal for {bucket_name}/{object_name} due to transient error: {err}"
+                                )))),
+                                HealObjectOutcome::Failed => (0, Err(err)),
                            },
                        }
                    };
@@ -839,11 +952,12 @@ impl ErasureSetHealer {
                });
            }

-            let mut completed_in_page = 0usize;
            while let Some((key, object, version_id, result)) = page_tasks.next().await {
+                let (object_size, result) = result;
                match result {
                    Ok(true) => {
                        *successful_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                        checkpoint_manager.add_processed_object(key).await?;
                        debug!(
                            target: "rustfs::heal::erasure_healer",
@@ -861,6 +975,7 @@ impl ErasureSetHealer {
                    Ok(false) => {
                        checkpoint_manager.add_processed_object(key).await?;
                        *successful_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                        debug!(
                            target: "rustfs::heal::erasure_healer",
                            event = EVENT_HEAL_ERASURE_OBJECT_STATE,
@@ -877,6 +992,7 @@ impl ErasureSetHealer {
                    Err(err @ Error::TaskCancelled) | Err(err @ Error::TaskTimeout) => return Err(err),
                    Err(Error::TransientSkip { message }) => {
                        *skipped_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                        checkpoint_manager.add_skipped_object(key).await?;
                        demote_to_debug_when!(!take_failure_log_sample(&mut transient_skip_samples_logged), warn, target: "rustfs::heal::erasure_healer", {
                            event = EVENT_HEAL_ERASURE_OBJECT_STATE,
@@ -893,6 +1009,7 @@ impl ErasureSetHealer {
                    }
                    Err(err) => {
                        *failed_objects += 1;
+                        bytes_processed = bytes_processed.saturating_add(object_size);
                        checkpoint_manager.add_failed_object(key).await?;
                        demote_to_debug_when!(!take_failure_log_sample(&mut failure_samples_logged), warn, target: "rustfs::heal::erasure_healer", {
                            event = EVENT_HEAL_ERASURE_OBJECT_STATE,
@@ -911,6 +1028,11 @@ impl ErasureSetHealer {

                *processed_objects += 1;
                completed_in_page += 1;
+                {
+                    let mut progress = self.progress.write().await;
+                    progress.set_current_object(Some(format!("{bucket}/{object}")));
+                    progress.update_progress(*processed_objects, *successful_objects, *failed_objects, bytes_processed);
+                }

                if completed_in_page.is_multiple_of(100) {
                    checkpoint_manager.update_position(bucket_index, page_resume_index).await?;
@@ -964,7 +1086,9 @@ impl ErasureSetHealer {
        progress.objects_scanned = state.total_objects;
        progress.objects_healed = state.successful_objects;
        progress.objects_failed = state.failed_objects;
-        progress.bytes_processed = 0; // set to 0 for now, can be extended later
+        progress.bytes_processed = 0; // Resume state tracks object counts, not byte counters.
+        progress.start_time = UNIX_EPOCH.checked_add(Duration::from_secs(state.start_time));
+        progress.last_update_time = UNIX_EPOCH.checked_add(Duration::from_secs(state.last_update));
        progress.set_current_object(state.current_object.clone());
    }
 }
@@ -1135,13 +1259,15 @@ mod resume_loop_tests {
    //! that emits programmable multi-version pages. These exercise the real loop
    //! logic (cursor seeding, per-version dedup, anti-loop guard, absence
    //! handling) — not merely a mock's own output.
-    use super::{ErasureSetHealer, target_outcomes_complete};
+    use super::{
+        ErasureSetHealer, NANOS_PER_SECOND, NEW_VERSION_SKIP_GRACE_SECS, should_skip_new_version, target_outcomes_complete,
+    };
    use crate::heal::progress::HealProgress;
    use crate::heal::resume::{
        CheckpointManager, RESUME_CHECKPOINT_FILE, ReplacementTargetIdentity, ResumeDeleteFailure, ResumeManager, ResumeUtils,
        compose_key,
    };
-    use crate::heal::storage::{DiskStatus, HealListItem, HealObjectInfo, HealStorageAPI};
+    use crate::heal::storage::{DiskStatus, HealLifecycleExpiryContext, HealListItem, HealObjectInfo, HealStorageAPI};
    use crate::heal::storage_api::status::BucketInfo;
    use crate::heal::{
        BUCKET_META_PREFIX, DiskOption, DiskStore, EcstoreError, Endpoint, HealDiskExt as _, RUSTFS_META_BUCKET, new_disk,
@@ -1149,7 +1275,7 @@ mod resume_loop_tests {
    use crate::{Error, Result};
    use rustfs_common::heal_channel::{HealOpts, HealRequestSource};
    use rustfs_madmin::heal_commands::{HealDriveInfo, HealResultItem, Infos};
-    use std::collections::{HashMap, VecDeque};
+    use std::collections::{HashMap, HashSet, VecDeque};
    use std::sync::atomic::{AtomicBool, Ordering};
    use std::sync::{Arc, Mutex};
    use tempfile::TempDir;
@@ -1160,10 +1286,37 @@ mod resume_loop_tests {
        HealListItem {
            name: name.to_string(),
            version_id: version.map(str::to_string),
+            mod_time_unix_nanos: None,
+            lifecycle_object_info: None,
            is_delete_marker: delete_marker,
        }
    }

+    fn item_with_mod_time(name: &str, version: Option<&str>, mod_time_secs: u64) -> HealListItem {
+        HealListItem {
+            name: name.to_string(),
+            version_id: version.map(str::to_string),
+            mod_time_unix_nanos: Some(i128::from(mod_time_secs).saturating_mul(NANOS_PER_SECOND)),
+            lifecycle_object_info: None,
+            is_delete_marker: false,
+        }
+    }
+
+    #[test]
+    fn new_version_filter_respects_grace_boundary() {
+        let started_at = 1_700_000_000;
+
+        assert!(!should_skip_new_version(None, started_at));
+        assert!(!should_skip_new_version(
+            Some(i128::from(started_at + NEW_VERSION_SKIP_GRACE_SECS).saturating_mul(NANOS_PER_SECOND)),
+            started_at,
+        ));
+        assert!(should_skip_new_version(
+            Some(i128::from(started_at + NEW_VERSION_SKIP_GRACE_SECS + 1).saturating_mul(NANOS_PER_SECOND)),
+            started_at,
+        ));
+    }
+
    #[test]
    fn target_outcomes_require_each_requested_endpoint_once_and_ok() {
        let result = HealResultItem {
@@ -1246,8 +1399,10 @@ mod resume_loop_tests {
        /// Target-specific physical readback evidence per `compose_key`; the
        /// fake models a healthy backend unless a test explicitly revokes it.
        replacement_commit_evidence: Mutex<HashMap<String, ReplacementCommitEvidence>>,
+        lifecycle_expired: Mutex<HashSet<String>>,
        /// every heal_object call recorded as (name, version_id)
        heal_calls: Mutex<Vec<(String, Option<String>)>>,
+        list_include_lifecycle_object_info: Mutex<Vec<bool>>,
        replacement_target_identity_sequences: Mutex<VecDeque<Vec<ReplacementTargetIdentity>>>,
        fail_listing: AtomicBool,
    }
@@ -1274,9 +1429,15 @@ mod resume_loop_tests {
                .unwrap()
                .insert(compose_key(name, version), ReplacementCommitEvidence::Error(message.to_string()));
        }
+        fn set_lifecycle_expired(&self, name: &str, version: Option<&str>) {
+            self.lifecycle_expired.lock().unwrap().insert(compose_key(name, version));
+        }
        fn calls(&self) -> Vec<(String, Option<String>)> {
            self.heal_calls.lock().unwrap().clone()
        }
+        fn list_include_lifecycle_object_info_calls(&self) -> Vec<bool> {
+            self.list_include_lifecycle_object_info.lock().unwrap().clone()
+        }
        fn fail_listing(&self) {
            self.fail_listing.store(true, Ordering::SeqCst);
        }
@@ -1330,6 +1491,23 @@ mod resume_loop_tests {
        async fn get_object_checksum(&self, _b: &str, _o: &str) -> Result<Option<String>> {
            Ok(None)
        }
+        async fn load_heal_lifecycle_expiry_context(&self, _bucket: &str) -> Result<Option<HealLifecycleExpiryContext>> {
+            Ok((!self.lifecycle_expired.lock().unwrap().is_empty()).then(HealLifecycleExpiryContext::test))
+        }
+        async fn enqueue_heal_lifecycle_expiry(
+            &self,
+            _context: &HealLifecycleExpiryContext,
+            _bucket: &str,
+            object: &str,
+            version_id: Option<&str>,
+            _object_info: Option<&HealObjectInfo>,
+        ) -> Result<bool> {
+            Ok(self
+                .lifecycle_expired
+                .lock()
+                .unwrap()
+                .contains(&compose_key(object, version_id)))
+        }
        async fn heal_object(
            &self,
            _bucket: &str,
@@ -1386,7 +1564,12 @@ mod resume_loop_tests {
            _bucket: &str,
            _prefix: &str,
            continuation_token: Option<&str>,
+            include_lifecycle_object_info: bool,
        ) -> Result<(Vec<HealListItem>, Option<String>, bool)> {
+            self.list_include_lifecycle_object_info
+                .lock()
+                .unwrap()
+                .push(include_lifecycle_object_info);
            if self.fail_listing.load(Ordering::SeqCst) {
                return Err(Error::other("injected listing failure"));
            }
@@ -1476,6 +1659,7 @@ mod resume_loop_tests {

    /// Drive one bucket heal pass; returns (processed, successful, failed, skipped, result).
    async fn run(env: &Env) -> (u64, u64, u64, u64, Result<()>) {
+        let state = env.resume.get_state().await;
        let mut current_object_index = 0usize;
        let mut processed = 0u64;
        let mut successful = 0u64;
@@ -1494,6 +1678,7 @@ mod resume_loop_tests {
                &mut skipped,
                &env.resume,
                &env.checkpoint,
+                state.start_time,
            )
            .await;
        (processed, successful, failed, skipped, result)
@@ -1559,6 +1744,7 @@ mod resume_loop_tests {
        let mut successful = 0;
        let mut failed = 0;
        let mut skipped = 0;
+        let started_at = env.resume.get_state().await.start_time;

        let error = healer
            .heal_bucket_with_resume(
@@ -1572,6 +1758,7 @@ mod resume_loop_tests {
                &mut skipped,
                &env.resume,
                &env.checkpoint,
+                started_at,
            )
            .await
            .expect_err("a remounted target must not begin a new page scan");
@@ -1641,6 +1828,109 @@ mod resume_loop_tests {
        assert_eq!(skipped, 0);
    }

+    #[tokio::test]
+    async fn erasure_set_progress_accumulates_healed_object_bytes() {
+        let env = make_env().await;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![item("first", Some("v1"), false), item("second", Some("v2"), false)],
+                next: None,
+                truncated: false,
+            },
+        );
+        env.storage.set_result(
+            "first",
+            Some("v1"),
+            HealResultItem {
+                object_size: 1024,
+                ..Default::default()
+            },
+        );
+        env.storage.set_result(
+            "second",
+            Some("v2"),
+            HealResultItem {
+                object_size: 2048,
+                ..Default::default()
+            },
+        );
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 2);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 2);
+        assert_eq!(progress.objects_failed, 0);
+        assert_eq!(progress.bytes_processed, 3072);
+        assert!(matches!(progress.current_object.as_deref(), Some("b/first" | "b/second")));
+    }
+
+    #[tokio::test]
+    async fn erasure_set_skips_versions_written_after_heal_started() {
+        let env = make_env().await;
+        let started_at = env.resume.get_state().await.start_time;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![
+                    item_with_mod_time("old", Some("v1"), started_at + NEW_VERSION_SKIP_GRACE_SECS),
+                    item_with_mod_time("new", Some("v2"), started_at + NEW_VERSION_SKIP_GRACE_SECS + 1),
+                ],
+                next: None,
+                truncated: false,
+            },
+        );
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 1);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        assert_eq!(env.storage.calls(), vec![("old".to_string(), Some("v1".to_string()))]);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.skipped_new_versions, 1);
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 1);
+        assert_eq!(progress.objects_failed, 0);
+    }
+
+    #[tokio::test]
+    async fn erasure_set_skips_versions_queued_for_lifecycle_expiry() {
+        let env = make_env().await;
+        env.storage.set_page(
+            None,
+            Page {
+                items: vec![item("expired", Some("v1"), false), item("kept", Some("v2"), false)],
+                next: None,
+                truncated: false,
+            },
+        );
+        env.storage.set_lifecycle_expired("expired", Some("v1"));
+
+        let (processed, successful, failed, skipped, result) = run(&env).await;
+
+        result.expect("page heal should succeed");
+        assert_eq!(processed, 2);
+        assert_eq!(successful, 1);
+        assert_eq!(failed, 0);
+        assert_eq!(skipped, 0);
+        assert_eq!(env.storage.calls(), vec![("kept".to_string(), Some("v2".to_string()))]);
+        assert_eq!(env.storage.list_include_lifecycle_object_info_calls(), vec![true]);
+        let progress = env.healer.progress.read().await;
+        assert_eq!(progress.skipped_ilm_expired, 1);
+        assert_eq!(progress.objects_scanned, 2);
+        assert_eq!(progress.objects_healed, 1);
+        assert_eq!(progress.objects_failed, 0);
+    }
+
    #[tokio::test]
    async fn bucket_listing_failure_does_not_mark_set_completed() {
        let env = make_env().await;
@@ -220,6 +220,11 @@ struct CompletedHealStatus {
    result_items: Vec<HealResultItem>,
    result_items_truncated: bool,
    completed_at: SystemTime,
+    /// Sequence-stamped retained window, archived with the completion so
+    /// incremental consumers keep their cursor across the transition (HS-06).
+    seqed_items: Vec<(u64, HealResultItem)>,
+    next_seq: u64,
+    min_seq: u64,
 }

 #[derive(Debug, Clone)]
@@ -240,6 +245,65 @@ pub struct HealTaskReport {
    pub result_items: Vec<HealResultItem>,
    pub result_items_truncated: bool,
    pub progress: Option<HealProgress>,
+    /// Cursor for incremental consumption: sequence number of the next item
+    /// to be produced. `0` on reports from sources without sequencing.
+    pub next_seq: u64,
+    /// Oldest sequence still retained (`0` together with `next_seq` when
+    /// sequencing is unavailable).
+    pub min_seq: u64,
+}
+
+/// Report from a live task, honoring the client's incremental cursor.
+async fn active_task_report(task: &HealTask, since: Option<u64>) -> HealTaskReport {
+    let window = task.get_result_items_since(since).await;
+    HealTaskReport {
+        status: task.get_status().await,
+        result_items: window.items,
+        // The legacy flag stays set once anything was evicted; a lagging
+        // incremental cursor additionally marks this response truncated so
+        // the client knows to restart from `min_seq`.
+        result_items_truncated: task.result_items_truncated() || window.lagged,
+        progress: Some(task.get_progress().await),
+        next_seq: window.next_seq,
+        min_seq: window.min_seq,
+    }
+}
+
+fn empty_task_report(status: HealTaskStatus) -> HealTaskReport {
+    HealTaskReport {
+        status,
+        result_items: Vec::new(),
+        result_items_truncated: false,
+        progress: None,
+        next_seq: 0,
+        min_seq: 0,
+    }
+}
+
+fn completed_task_report(completed: &CompletedHealStatus, since: Option<u64>) -> HealTaskReport {
+    let mut lagged = false;
+    let result_items = match since {
+        None => completed.result_items.clone(),
+        Some(cursor) => {
+            if cursor + 1 < completed.min_seq {
+                lagged = true;
+            }
+            completed
+                .seqed_items
+                .iter()
+                .filter(|(seq, _)| *seq > cursor)
+                .map(|(_, item)| item.clone())
+                .collect()
+        }
+    };
+    HealTaskReport {
+        status: completed.status.clone(),
+        result_items,
+        result_items_truncated: completed.result_items_truncated || lagged,
+        progress: None,
+        next_seq: completed.next_seq,
+        min_seq: completed.min_seq,
+    }
 }

 #[derive(Debug, Clone, Copy, Default, PartialEq, Eq, serde::Deserialize, serde::Serialize)]
@@ -270,6 +334,8 @@ pub struct HealSourceCounts {
    pub auto_heal: u64,
    pub internal: u64,
    pub read_repair: u64,
+    #[serde(default)]
+    pub mrf: u64,
 }

 impl HealSourceCounts {
@@ -280,6 +346,7 @@ impl HealSourceCounts {
            HealRequestSource::AutoHeal => self.auto_heal += 1,
            HealRequestSource::Internal => self.internal += 1,
            HealRequestSource::ReadRepair => self.read_repair += 1,
+            HealRequestSource::Mrf => self.mrf += 1,
        }
    }
 }
@@ -528,6 +595,11 @@ impl PriorityHealQueue {
        self.dedup_keys.contains_key(&key)
    }

+    /// Iterate queued requests (used by the admin overlap check).
+    fn requests(&self) -> impl Iterator<Item = &HealRequest> {
+        self.heap.iter().map(|item| &item.request)
+    }
+
    fn contains_request_id(&self, request_id: &str) -> bool {
        self.heap.iter().any(|item| item.request.id == request_id)
    }
@@ -686,6 +758,80 @@ fn recoverable_heal_retry_delay(retry_attempt: u32) -> Duration {
 }

 /// Heal config
+/// HS-06 admin overlap policy.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
+pub enum HealOverlapPolicy {
+    /// Default: overlapping admin starts merge into the existing task
+    /// (today's dedup semantics).
+    #[default]
+    Merge,
+    /// Return a typed already-running / overlapping-paths rejection like
+    /// madmin's ErrHealAlreadyRunning / ErrHealOverlappingPaths.
+    MinioError,
+}
+
+/// Path view of a heal type for overlap comparison: a bucket plus a
+/// prefix/object path inside it (`None` bucket = cluster-wide, overlaps
+/// everything).
+fn heal_type_path_view(heal_type: &HealType) -> (Option<&str>, &str) {
+    match heal_type {
+        HealType::Cluster => (None, ""),
+        HealType::Bucket { bucket } => (Some(bucket), ""),
+        HealType::Prefix { bucket, prefix } => (Some(bucket), prefix),
+        HealType::Object { bucket, object, .. }
+        | HealType::Metadata { bucket, object }
+        | HealType::ECDecode { bucket, object, .. } => (Some(bucket), object),
+        // MRF/MetaPath heal keys on a meta path; treat the whole set of
+        // buckets as one namespace so it only overlaps itself exactly.
+        HealType::MRF { meta_path } => (Some("\u{0}mrf"), meta_path),
+        // Erasure-set heal: the set id is the overlap dimension.
+        HealType::ErasureSet { set_disk_id, .. } => (Some("\u{0}set"), set_disk_id),
+    }
+}
+
+/// How two heal paths relate for the admin overlap check (HS-06).
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum OverlapVerdict {
+    /// Distinct targets: no conflict.
+    Disjoint,
+    /// Same target: an identical heal is already in flight.
+    SameTarget,
+    /// One target contains the other.
+    Overlapping,
+}
+
+fn prefix_paths_overlap(a: &str, b: &str) -> OverlapVerdict {
+    if a == b {
+        return OverlapVerdict::SameTarget;
+    }
+    if a.is_empty() || b.is_empty() || a.starts_with(b) || b.starts_with(a) {
+        return OverlapVerdict::Overlapping;
+    }
+    OverlapVerdict::Disjoint
+}
+
+fn heal_types_overlap(left: &HealType, right: &HealType) -> OverlapVerdict {
+    let (left_bucket, left_path) = heal_type_path_view(left);
+    let (right_bucket, right_path) = heal_type_path_view(right);
+    match (left_bucket, right_bucket) {
+        // Cluster-wide overlaps everything (but an exact cluster match is
+        // SameTarget).
+        (None, _) | (_, None) => {
+            if matches!(left, HealType::Cluster) && matches!(right, HealType::Cluster) {
+                OverlapVerdict::SameTarget
+            } else {
+                OverlapVerdict::Overlapping
+            }
+        }
+        (Some(lb), Some(rb)) => {
+            if lb != rb {
+                return OverlapVerdict::Disjoint;
+            }
+            prefix_paths_overlap(left_path, right_path)
+        }
+    }
+}
+
 #[derive(Debug, Clone)]
 pub struct HealConfig {
    /// Whether to enable auto heal
@@ -706,6 +852,9 @@ pub struct HealConfig {
    pub low_priority_drop_when_full: bool,
    /// Whether notify-driven scheduler wakeups are enabled.
    pub event_driven_scheduler_enable: bool,
+    /// How admin heal starts behave on path overlap (HS-06): merge into the
+    /// existing task (default) or return a typed already-running rejection.
+    pub overlap_policy: HealOverlapPolicy,
    /// Whether per-set bulkhead scheduling is enabled.
    pub set_bulkhead_enable: bool,
    /// Whether erasure-set page parallelism is enabled.
@@ -754,6 +903,14 @@ impl Default for HealConfig {
            rustfs_config::ENV_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE,
            rustfs_config::DEFAULT_HEAL_EVENT_DRIVEN_SCHEDULER_ENABLE,
        );
+        let overlap_policy =
+            match rustfs_utils::get_env_str(rustfs_config::ENV_HEAL_OVERLAP_POLICY, rustfs_config::DEFAULT_HEAL_OVERLAP_POLICY)
+                .to_lowercase()
+                .as_str()
+            {
+                "minio_error" => HealOverlapPolicy::MinioError,
+                _ => HealOverlapPolicy::Merge,
+            };
        let set_bulkhead_enable = rustfs_utils::get_env_bool(
            rustfs_config::ENV_HEAL_SET_BULKHEAD_ENABLE,
            rustfs_config::DEFAULT_HEAL_SET_BULKHEAD_ENABLE,
@@ -790,6 +947,7 @@ impl Default for HealConfig {
            low_priority_merge_enable,
            low_priority_drop_when_full,
            event_driven_scheduler_enable,
+            overlap_policy,
            set_bulkhead_enable,
            page_parallel_enable,
            mainline_throttle_enable,
@@ -1756,6 +1914,50 @@ impl HealManager {
        request: HealRequest,
        preserve_alias: bool,
    ) -> Result<HealAdmissionReceipt> {
+        // HS-06 forceStart semantics (admin only): MinIO stops the old task
+        // first and then starts the new one. Cancel any active admin task
+        // overlapping this request's path before entering admission, so the
+        // fresh task is never merged into the one being replaced.
+        if request.source == HealRequestSource::Admin && request.force_start {
+            let overlapping: Vec<String> = {
+                let active_heals = self.active_heals.lock().await;
+                active_heals
+                    .iter()
+                    .filter(|(task_id, task)| {
+                        task.source == HealRequestSource::Admin
+                            && heal_types_overlap(&request.heal_type, &task.heal_type) != OverlapVerdict::Disjoint
+                            && *task_id != &request.id
+                    })
+                    .map(|(task_id, _)| task_id.clone())
+                    .collect()
+            };
+            for task_id in overlapping {
+                match self.cancel_task(&task_id).await {
+                    Ok(_) => info!(
+                        target: "rustfs::heal::manager",
+                        event = EVENT_HEAL_QUEUE_ADMISSION,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_MANAGER,
+                        request_id = %request.id,
+                        cancelled_task_id = %task_id,
+                        result = "force_start_cancelled_overlap",
+                        "Admin forceStart cancelled an overlapping heal task"
+                    ),
+                    Err(err) => warn!(
+                        target: "rustfs::heal::manager",
+                        event = EVENT_HEAL_QUEUE_ADMISSION,
+                        component = LOG_COMPONENT_HEAL,
+                        subsystem = LOG_SUBSYSTEM_MANAGER,
+                        request_id = %request.id,
+                        cancelled_task_id = %task_id,
+                        error = %err,
+                        result = "force_start_cancel_failed",
+                        "Admin forceStart failed to cancel an overlapping heal task"
+                    ),
+                }
+            }
+        }
+
        let config = self.config.read().await;
        let dedup_key = PriorityHealQueue::make_dedup_key(&request);

@@ -1778,7 +1980,15 @@ impl HealManager {
                .or_else(|| retrying_heal_for_dedup_key(&retrying_heals, &dedup_key).map(|(task_id, _)| (task_id, "retrying")))
        });
        if let Some((merged_task_id, duplicate_state)) = duplicate.flatten() {
-            let admission = Self::duplicate_admission_for_request(&request, &config);
+            // HS-06: under the minio_error overlap policy an exact duplicate
+            // admin start reports the typed AlreadyRunning rejection instead
+            // of the silent merge (MinIO's ErrHealAlreadyRunning).
+            let admission =
+                if request.source == HealRequestSource::Admin && config.overlap_policy == HealOverlapPolicy::MinioError {
+                    HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning)
+                } else {
+                    Self::duplicate_admission_for_request(&request, &config)
+                };
            drop(retrying_heals);
            drop(queue);
            drop(active_heals);
@@ -1824,6 +2034,62 @@ impl HealManager {
            });
        }

+        // HS-06 typed overlap rejection (admin only, minio_error policy):
+        // paths containing or contained by an active/queued task reject with
+        // AlreadyRunning / OverlappingPaths instead of merging. Exact
+        // duplicates already merged above; scanner/autoheal/read-repair
+        // sources never take this path.
+        if request.source == HealRequestSource::Admin && config.overlap_policy == HealOverlapPolicy::MinioError {
+            let mut rejection = None;
+            for (task_id, task) in active_heals.iter() {
+                match heal_types_overlap(&request.heal_type, &task.heal_type) {
+                    OverlapVerdict::SameTarget => {
+                        rejection = Some((HealAdmissionDropReason::AlreadyRunning, task_id.clone()));
+                        break;
+                    }
+                    OverlapVerdict::Overlapping => {
+                        rejection = Some((HealAdmissionDropReason::OverlappingPaths, task_id.clone()));
+                    }
+                    OverlapVerdict::Disjoint => {}
+                }
+            }
+            if rejection.is_none() {
+                for queued in queue.requests() {
+                    match heal_types_overlap(&request.heal_type, &queued.heal_type) {
+                        OverlapVerdict::SameTarget => {
+                            rejection = Some((HealAdmissionDropReason::AlreadyRunning, queued.id.clone()));
+                            break;
+                        }
+                        OverlapVerdict::Overlapping => {
+                            rejection = Some((HealAdmissionDropReason::OverlappingPaths, queued.id.clone()));
+                        }
+                        OverlapVerdict::Disjoint => {}
+                    }
+                }
+            }
+            if let Some((reason, overlap_task_id)) = rejection {
+                drop(retrying_heals);
+                drop(queue);
+                drop(active_heals);
+                Self::record_admission_metric(request.source, HealAdmissionResult::Dropped(reason), "overlap_rejected");
+                warn!(
+                    target: "rustfs::heal::manager",
+                    event = EVENT_HEAL_QUEUE_ADMISSION,
+                    component = LOG_COMPONENT_HEAL,
+                    subsystem = LOG_SUBSYSTEM_MANAGER,
+                    request_id = %request.id,
+                    overlap_task_id = %overlap_task_id,
+                    reason = reason.as_str(),
+                    result = "overlap_rejected",
+                    "Admin heal start rejected by overlap policy"
+                );
+                return Ok(HealAdmissionReceipt {
+                    result: HealAdmissionResult::Dropped(reason),
+                    task_id: overlap_task_id,
+                });
+            }
+        }
+
        let mut task_id = request.id.clone();
        let admission = Self::admit_request_to_queue(&mut queue, request, &config, "submit");
        if admission == HealAdmissionResult::Merged
@@ -1896,28 +2162,25 @@ impl HealManager {
    }

    pub async fn get_task_report(&self, task_id: &str) -> Result<HealTaskReport> {
+        self.get_task_report_since(task_id, None).await
+    }
+
+    /// Incremental variant of [`Self::get_task_report`] (HS-06): `since` is
+    /// the client's last seen sequence number; `None` keeps the legacy
+    /// full-snapshot semantics.
+    pub async fn get_task_report_since(&self, task_id: &str, since: Option<u64>) -> Result<HealTaskReport> {
        let canonical_task_id = self.canonical_task_id(task_id).await;
        {
            let active_heals = self.active_heals.lock().await;
            if let Some(task) = active_heals.get(&canonical_task_id) {
-                return Ok(HealTaskReport {
-                    status: task.get_status().await,
-                    result_items: task.get_result_items().await,
-                    result_items_truncated: task.result_items_truncated(),
-                    progress: Some(task.get_progress().await),
-                });
+                return Ok(active_task_report(task, since).await);
            }
        }

        {
            let retrying_heals = self.retrying_heals.lock().await;
            if let Some(retrying) = retrying_heals.get(&canonical_task_id) {
-                return Ok(HealTaskReport {
-                    status: retrying.status(),
-                    result_items: Vec::new(),
-                    result_items_truncated: false,
-                    progress: None,
-                });
+                return Ok(empty_task_report(retrying.status()));
            }
        }

@@ -1927,36 +2190,21 @@ impl HealManager {
            if let Some(completed) = completed_heals.get(&canonical_task_id)
                && completed_status_is_retrying(&completed.status)
            {
-                return Ok(HealTaskReport {
-                    status: completed.status.clone(),
-                    result_items: completed.result_items.clone(),
-                    result_items_truncated: completed.result_items_truncated,
-                    progress: None,
-                });
+                return Ok(completed_task_report(completed, since));
            }
        }

        {
            let queue = self.heal_queue.lock().await;
            if queue.contains_request_id(&canonical_task_id) {
-                return Ok(HealTaskReport {
-                    status: HealTaskStatus::Pending,
-                    result_items: Vec::new(),
-                    result_items_truncated: false,
-                    progress: None,
-                });
+                return Ok(empty_task_report(HealTaskStatus::Pending));
            }
        }

        let mut completed_heals = self.completed_heals.lock().await;
        prune_completed_heal_statuses(&mut completed_heals);
        if let Some(completed) = completed_heals.get(&canonical_task_id) {
-            return Ok(HealTaskReport {
-                status: completed.status.clone(),
-                result_items: completed.result_items.clone(),
-                result_items_truncated: completed.result_items_truncated,
-                progress: None,
-            });
+            return Ok(completed_task_report(completed, since));
        }

        Err(Error::TaskNotFound {
@@ -1965,18 +2213,23 @@ impl HealManager {
    }

    pub async fn get_task_report_for_path(&self, heal_path: &str, task_id: &str) -> Result<HealTaskReport> {
+        self.get_task_report_for_path_since(heal_path, task_id, None).await
+    }
+
+    /// Incremental variant of [`Self::get_task_report_for_path`] (HS-06).
+    pub async fn get_task_report_for_path_since(
+        &self,
+        heal_path: &str,
+        task_id: &str,
+        since: Option<u64>,
+    ) -> Result<HealTaskReport> {
        let canonical_task_id = self.canonical_task_id(task_id).await;
        {
            let active_heals = self.active_heals.lock().await;
            if let Some(task) = active_heals.get(&canonical_task_id)
                && heal_type_matches_path(&task.heal_type, heal_path)
            {
-                return Ok(HealTaskReport {
-                    status: task.get_status().await,
-                    result_items: task.get_result_items().await,
-                    result_items_truncated: task.result_items_truncated(),
-                    progress: Some(task.get_progress().await),
-                });
+                return Ok(active_task_report(task, since).await);
            }
        }

@@ -1985,12 +2238,7 @@ impl HealManager {
            if let Some(retrying) = retrying_heals.get(&canonical_task_id)
                && heal_type_matches_path(&retrying.request.heal_type, heal_path)
            {
-                return Ok(HealTaskReport {
-                    status: retrying.status(),
-                    result_items: Vec::new(),
-                    result_items_truncated: false,
-                    progress: None,
-                });
+                return Ok(empty_task_report(retrying.status()));
            }
        }

@@ -2001,24 +2249,14 @@ impl HealManager {
                && heal_type_matches_path(&completed.heal_type, heal_path)
                && completed_status_is_retrying(&completed.status)
            {
-                return Ok(HealTaskReport {
-                    status: completed.status.clone(),
-                    result_items: completed.result_items.clone(),
-                    result_items_truncated: completed.result_items_truncated,
-                    progress: None,
-                });
+                return Ok(completed_task_report(completed, since));
            }
        }

        {
            let queue = self.heal_queue.lock().await;
            if queue.contains_request_id_matching_path(&canonical_task_id, heal_path) {
-                return Ok(HealTaskReport {
-                    status: HealTaskStatus::Pending,
-                    result_items: Vec::new(),
-                    result_items_truncated: false,
-                    progress: None,
-                });
+                return Ok(empty_task_report(HealTaskStatus::Pending));
            }
        }

@@ -2028,12 +2266,7 @@ impl HealManager {
            if let Some(completed) = completed_heals.get(&canonical_task_id)
                && heal_type_matches_path(&completed.heal_type, heal_path)
            {
-                return Ok(HealTaskReport {
-                    status: completed.status.clone(),
-                    result_items: completed.result_items.clone(),
-                    result_items_truncated: completed.result_items_truncated,
-                    progress: None,
-                });
+                return Ok(completed_task_report(completed, since));
            }
        }

@@ -2385,8 +2618,27 @@ impl HealManager {
            snapshot.objects_scanned = snapshot.objects_scanned.saturating_add(progress.objects_scanned);
            snapshot.objects_healed = snapshot.objects_healed.saturating_add(progress.objects_healed);
            snapshot.objects_failed = snapshot.objects_failed.saturating_add(progress.objects_failed);
+            snapshot.skipped_new_versions = snapshot.skipped_new_versions.saturating_add(progress.skipped_new_versions);
+            snapshot.skipped_ilm_expired = snapshot.skipped_ilm_expired.saturating_add(progress.skipped_ilm_expired);
+            snapshot.objects_total_count = snapshot.objects_total_count.saturating_add(progress.objects_total_count);
+            snapshot.objects_total_size = snapshot.objects_total_size.saturating_add(progress.objects_total_size);
            snapshot.bytes_processed = snapshot.bytes_processed.saturating_add(progress.bytes_processed);
+            snapshot.start_time = match (snapshot.start_time, progress.start_time) {
+                (Some(current), Some(next)) => Some(current.min(next)),
+                (None, next) => next,
+                (current, None) => current,
+            };
+            snapshot.last_update_time = match (snapshot.last_update_time, progress.last_update_time) {
+                (Some(current), Some(next)) => Some(current.max(next)),
+                (None, next) => next,
+                (current, None) => current,
+            };
+            if progress.current_object.is_some() {
+                snapshot.current_object = progress.current_object;
+            }
        }
+        snapshot.refresh_progress_percentage();
+        snapshot.refresh_estimated_completion_time();
        Some(snapshot)
    }

@@ -3208,12 +3460,17 @@ impl HealManager {
                        } else {
                            completed_task.get_status().await
                        };
+                        let completed_progress = completed_task.get_progress().await;
+                        let final_window = completed_task.get_result_items_since(None).await;
                        let completed_status_entry = CompletedHealStatus {
                            heal_type: completed_task.heal_type.clone(),
                            status: completed_status.clone(),
-                            result_items: completed_task.get_result_items().await,
+                            result_items: final_window.items.clone(),
                            result_items_truncated: completed_task.result_items_truncated(),
                            completed_at: SystemTime::now(),
+                            seqed_items: completed_task.get_seqed_result_items().await,
+                            next_seq: final_window.next_seq,
+                            min_seq: final_window.min_seq,
                        };
                        let mut completed_heals_guard = completed_heals_clone.lock().await;
                        prune_completed_heal_statuses(&mut completed_heals_guard);
@@ -3223,6 +3480,7 @@ impl HealManager {
                        match completed_status {
                            HealTaskStatus::Completed => {
                                stats.update_task_completion(true);
+                                stats.add_healed_objects(completed_progress.objects_healed, completed_progress.bytes_processed);
                            }
                            HealTaskStatus::Retrying { .. } => {}
                            _ => {
@@ -3749,6 +4007,7 @@ mod tests {
            _bucket: &str,
            _prefix: &str,
            _continuation_token: Option<&str>,
+            _include_lifecycle_object_info: bool,
        ) -> Result<(Vec<crate::heal::storage::HealListItem>, Option<String>, bool)> {
            Ok((Vec::new(), None, false))
        }
@@ -4983,6 +5242,9 @@ mod tests {
                },
                result_items: Vec::new(),
                result_items_truncated: false,
+                seqed_items: Vec::new(),
+                next_seq: 0,
+                min_seq: 0,
                completed_at: SystemTime::now(),
            },
        );
@@ -5264,6 +5526,136 @@ mod tests {
        assert_eq!(snapshot.queued_by_source.internal, 0);
    }

+    // HS-06 (backlog#1870): overlap policy + forceStart semantics.
+    fn manager_with_policy(policy: HealOverlapPolicy) -> HealManager {
+        let storage: Arc<dyn HealStorageAPI> = Arc::new(MockStorage);
+        HealManager::new(
+            storage,
+            Some(HealConfig {
+                overlap_policy: policy,
+                ..Default::default()
+            }),
+        )
+    }
+
+    fn admin_prefix_request(bucket: &str, prefix: &str) -> HealRequest {
+        let mut request = HealRequest::new(
+            HealType::Prefix {
+                bucket: bucket.to_string(),
+                prefix: prefix.to_string(),
+            },
+            HealOptions::default(),
+            HealPriority::Normal,
+        );
+        request.source = HealRequestSource::Admin;
+        request
+    }
+
+    async fn insert_active_task(manager: &HealManager, request: HealRequest) -> String {
+        let task = Arc::new(HealTask::from_request(request, manager.storage.clone()));
+        let task_id = task.id.clone();
+        manager.active_heals.lock().await.insert(task_id.clone(), task);
+        task_id
+    }
+
+    #[tokio::test]
+    async fn overlap_policy_minio_error_rejects_same_and_containing_paths() {
+        let manager = manager_with_policy(HealOverlapPolicy::MinioError);
+        insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await;
+
+        // Same target: typed AlreadyRunning.
+        let same = manager
+            .submit_heal_request(admin_prefix_request("bucket-a", "logs/"))
+            .await
+            .expect("admission must decide");
+        assert_eq!(
+            same,
+            HealAdmissionResult::Dropped(HealAdmissionDropReason::AlreadyRunning),
+            "an identical target must reject with already-running"
+        );
+
+        // Contained path: typed OverlappingPaths.
+        let nested = manager
+            .submit_heal_request(admin_prefix_request("bucket-a", "logs/app/"))
+            .await
+            .expect("admission must decide");
+        assert_eq!(
+            nested,
+            HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths),
+            "a path inside the active task's path must reject with overlapping-paths"
+        );
+
+        // Containing path (bucket-wide vs nested active): also overlapping.
+        let wide = manager
+            .submit_heal_request(admin_prefix_request("bucket-a", ""))
+            .await
+            .expect("admission must decide");
+        assert_eq!(
+            wide,
+            HealAdmissionResult::Dropped(HealAdmissionDropReason::OverlappingPaths),
+            "a bucket-wide start overlapping a nested active heal must reject"
+        );
+
+        // Disjoint bucket: unaffected.
+        let disjoint = manager
+            .submit_heal_request(admin_prefix_request("bucket-b", "logs/"))
+            .await
+            .expect("admission must decide");
+        assert_eq!(disjoint, HealAdmissionResult::Accepted);
+    }
+
+    #[tokio::test]
+    async fn overlap_policy_default_merge_keeps_today_semantics() {
+        let manager = manager_with_policy(HealOverlapPolicy::Merge);
+        insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await;
+
+        // Different-dedup-key overlap still merges under the default policy:
+        // the nested path dedups to its own key but nothing rejects it.
+        let nested = manager
+            .submit_heal_request(admin_prefix_request("bucket-a", "logs/app/"))
+            .await
+            .expect("admission must decide");
+        assert_eq!(nested, HealAdmissionResult::Accepted, "default policy must not reject overlaps");
+
+        // Non-admin sources never get overlap rejections even under minio_error.
+        let manager = manager_with_policy(HealOverlapPolicy::MinioError);
+        insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await;
+        let mut scanner_request = admin_prefix_request("bucket-a", "logs/app/");
+        scanner_request.source = HealRequestSource::Scanner;
+        let admitted = manager
+            .submit_heal_request(scanner_request)
+            .await
+            .expect("admission must decide");
+        assert_eq!(admitted, HealAdmissionResult::Accepted, "scanner sources must never be overlap-rejected");
+    }
+
+    #[tokio::test]
+    async fn admin_force_start_cancels_overlapping_active_task_first() {
+        let manager = manager_with_policy(HealOverlapPolicy::Merge);
+        let old_id = insert_active_task(&manager, admin_prefix_request("bucket-a", "logs/")).await;
+
+        let mut replacement = admin_prefix_request("bucket-a", "logs/");
+        replacement.force_start = true;
+        let receipt = manager
+            .submit_heal_request_with_receipt(replacement)
+            .await
+            .expect("force-start submission must decide");
+
+        assert!(receipt.result.is_admitted(), "the new task must be admitted (Accepted or Merged)");
+        let old_task_gone = {
+            let active_heals = manager.active_heals.lock().await;
+            !active_heals.contains_key(&old_id)
+        };
+        assert!(
+            old_task_gone,
+            "the overlapping admin task must be cancelled (removed from the active table) before the new one starts"
+        );
+        assert!(
+            matches!(manager.get_task_status(&old_id).await, Err(Error::TaskNotFound { .. })),
+            "a cancelled task must no longer resolve as an active heal"
+        );
+    }
+
    #[tokio::test]
    async fn test_operations_snapshot_counts_active_by_source_and_priority() {
        let storage: Arc<dyn HealStorageAPI> = Arc::new(MockStorage);
@@ -5396,6 +5788,8 @@ mod tests {
        ));
        {
            let mut progress = first.progress.write().await;
+            progress.start_time = Some(SystemTime::now() - Duration::from_secs(20));
+            progress.set_total_baseline(12, 8192);
            progress.update_progress(7, 3, 1, 4096);
        }

@@ -5405,6 +5799,8 @@ mod tests {
        ));
        {
            let mut progress = second.progress.write().await;
+            progress.start_time = Some(SystemTime::now() - Duration::from_secs(10));
+            progress.set_total_baseline(8, 4096);
            progress.update_progress(11, 5, 2, 2048);
        }

@@ -5419,7 +5815,11 @@ mod tests {
        assert_eq!(progress.objects_scanned, 18);
        assert_eq!(progress.objects_healed, 8);
        assert_eq!(progress.objects_failed, 3);
+        assert_eq!(progress.objects_total_count, 20);
+        assert_eq!(progress.objects_total_size, 12288);
        assert_eq!(progress.bytes_processed, 6144);
+        assert!((progress.progress_percentage - 50.0).abs() < 0.001);
+        assert!(progress.estimated_completion_time.is_some());
    }

    #[tokio::test]
@@ -5558,6 +5958,9 @@ mod tests {
                status: HealTaskStatus::Completed,
                result_items: Vec::new(),
                result_items_truncated: false,
+                seqed_items: Vec::new(),
+                next_seq: 0,
+                min_seq: 0,
                completed_at: SystemTime::now(),
            },
        );
@@ -5592,6 +5995,9 @@ mod tests {
                    ..Default::default()
                }],
                result_items_truncated: true,
+                seqed_items: Vec::new(),
+                next_seq: 0,
+                min_seq: 0,
                completed_at: SystemTime::now(),
            },
        );
--- a/Show More
+++ b/Show More