fix(scanner): correct observational usage arguments

CLAUDE.md

@@ -30,8 +30,7 @@ make build-docker BUILD_OS=ubuntu22.04
 - Crate membership: `Cargo.toml` `[workspace].members`
 - Architecture, layering, crate map: [ARCHITECTURE.md](ARCHITECTURE.md)
 - Migration guardrails & readiness contracts: [docs/architecture/](docs/architecture/README.md)
-- CI workflow steps: `.github/workflows/`; event, timeout, and required-status
-  matrix: [docs/testing/ci-gates.md](docs/testing/ci-gates.md)
+- CI gates: `.github/workflows/ci.yml` (source of truth; never copy its steps into docs)
 - Test-layer taxonomy, per-layer entry commands, serial/nextest rules, flake
   policy: [docs/testing/README.md](docs/testing/README.md)
 - Tier/ILM transition debugging (xl.meta inspection, versionId tracing):

CONTRIBUTING.md

@@ -70,8 +70,6 @@ make pre-pr
 
 > For the full test-layer taxonomy (unit / ecstore black-box / e2e / s3s-e2e / S3 compatibility / chaos / fuzz / bench), each layer's entry command, the naming conventions the migration gate depends on, and the serial/nextest rules, see [docs/testing/README.md](docs/testing/README.md).
 
-> For the event, timeout, required-status, and local reproduction matrix, see [docs/testing/ci-gates.md](docs/testing/ci-gates.md).
-
 ### 🔒 Automated Pre-commit Hooks
 #### What `make pre-commit` and `make pre-pr` actually run
 

Cargo.lock

@@ -1858,9 +1858,9 @@ dependencies = [
 
 [[package]]
 name = "cc"
-version = "1.4.4"
+version = "1.4.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0ad534f4357a5264cce5019c989cf66a4f0dc4e0d1b1d15f8aacec0ff7360273"
+checksum = "509591b7bcd67f4ef775afad7662703b4935daaa6ec0e5605cfb1090b32a2b6d"
 dependencies = [
  "find-msvc-tools",
  "jobserver",
@@ -2522,6 +2522,12 @@ dependencies = [
  "subtle",
 ]
 
+[[package]]
+name = "cty"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b365fabc795046672053e29c954733ec3b05e4be654ab130fe8f1f94d7051f35"
+
 [[package]]
 name = "curve25519-dalek"
 version = "4.1.3"
@@ -5982,6 +5988,15 @@ version = "0.2.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6d2cec3eae94f9f509c767b45932f1ada8350c4bdb85af2fcab4a3c14807981"
 
+[[package]]
+name = "libmimalloc-sys"
+version = "0.1.49"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "cc",
+ "cty",
+]
+
 [[package]]
 name = "libredox"
 version = "0.1.20"
@@ -6382,6 +6397,14 @@ dependencies = [
  "synstructure 0.13.2",
 ]
 
+[[package]]
+name = "mimalloc"
+version = "0.1.52"
+source = "git+https://github.com/xonatius/mimalloc_rust.git?rev=6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11#6d4c41bb10c6d9da1d1b6f07b38c4cc051667f11"
+dependencies = [
+ "libmimalloc-sys",
+]
+
 [[package]]
 name = "mime"
 version = "0.3.17"
@@ -9139,11 +9162,13 @@ dependencies = [
  "insta",
  "jiff",
  "libc",
+ "libmimalloc-sys",
  "libsystemd",
  "matchit 0.9.2",
  "md-5 0.11.0",
  "metrics",
  "metrics-util",
+ "mimalloc",
  "mime_guess",
  "opentelemetry",
  "opentelemetry_sdk",
@@ -9179,8 +9204,6 @@ dependencies = [
  "rustfs-lock",
  "rustfs-log-analyzer",
  "rustfs-madmin",
- "rustfs-mimalloc",
- "rustfs-mimalloc-sys",
  "rustfs-notify",
  "rustfs-object-capacity",
  "rustfs-object-data-cache",
@@ -9852,24 +9875,6 @@ dependencies = [
  "tokio",
 ]
 
-[[package]]
-name = "rustfs-mimalloc"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a406f4aa07084301d485beec873af6dccc8e3f8762da244743df92038b1db1a6"
-dependencies = [
- "rustfs-mimalloc-sys",
-]
-
-[[package]]
-name = "rustfs-mimalloc-sys"
-version = "0.5.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c3051b819175f58445d4c369a72f0ab88149f3885ba8bea2aff3be01f53fe7cd"
-dependencies = [
- "cc",
-]
-
 [[package]]
 name = "rustfs-notify"
 version = "1.0.0-rc.3"

Cargo.toml

@@ -350,8 +350,8 @@ russh-sftp = "2.4.0"
 dav-server = "0.11.0"
 
 # Performance Analysis and Memory Profiling
-rustfs-mimalloc = { version = "0.5.0" }
-rustfs-mimalloc-sys = { version = "0.5.0" }
+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 = "0.23.3", default-features = false }
 # Snapshot testing for output format regression detection
 insta = { version = "1.48" }

crates/data-usage/src/data_usage.rs

@@ -236,6 +236,15 @@ pub struct DataUsageInfo {
     /// without relying on synchronized clocks.
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub usage_snapshot_authoritative_baseline: Option<DataUsageSnapshotIdentity>,
+    /// Per-set freshness for an observational aggregate.  A set entry is
+    /// never sufficient to make the aggregate authoritative; it only records
+    /// which last-known-good generation contributed to the view.
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    pub usage_snapshot_set_states: Vec<DataUsageSnapshotSetState>,
+    /// An observational view may contain only the sets that completed this
+    /// cycle (or retained a compatible last-known-good cache).
+    #[serde(default)]
+    pub usage_snapshot_partial: bool,
     /// Deprecated kept here for backward compatibility reasons
     pub bucket_sizes: HashMap<String, u64>,
     /// Per-disk snapshot information when available
@@ -252,6 +261,22 @@ pub struct DataUsageSnapshotIdentity {
     pub scanner_epoch: Option<u64>,
 }
 
+#[derive(Clone, Debug, Default, Serialize, Deserialize, PartialEq, Eq)]
+pub struct DataUsageSnapshotSetState {
+    pub pool_index: u64,
+    pub set_index: u64,
+    #[serde(default)]
+    pub scanner_cycle: Option<u64>,
+    #[serde(default)]
+    pub scanner_epoch: Option<u64>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub scan_plan_digest: Option<[u8; 32]>,
+    #[serde(default)]
+    pub complete: bool,
+    #[serde(default)]
+    pub tombstone: bool,
+}
+
 impl DataUsageInfo {
     pub fn snapshot_identity(&self) -> DataUsageSnapshotIdentity {
         DataUsageSnapshotIdentity {
@@ -291,7 +316,7 @@ pub fn data_usage_snapshot_is_newer(candidate: &DataUsageInfo, baseline: &DataUs
 /// rollback delete/recreate fences the previous bucket incarnation too.
 pub fn observed_data_usage_is_newer(observed: &DataUsageInfo, authoritative: &DataUsageInfo) -> bool {
     observed.usage_snapshot_converged == Some(false)
-        && observed.is_complete_bucket_usage_snapshot()
+        && (observed.is_complete_bucket_usage_snapshot() || observed.is_valid_partial_snapshot())
         && observed.usage_snapshot_authoritative_baseline.as_ref() == Some(&authoritative.snapshot_identity())
         && data_usage_snapshot_is_newer(observed, authoritative)
 }
@@ -1436,6 +1461,39 @@ impl DataUsageInfo {
             && u64::try_from(self.buckets_usage.len()).ok() == Some(self.buckets_count)
     }
 
+    /// Validate provenance before an observational view can be selected for
+    /// admin display. Partial data is accepted only with unique set states,
+    /// a plan digest for every state, and at least one usable generation.
+    pub fn is_valid_partial_snapshot(&self) -> bool {
+        if !self.usage_snapshot_partial
+            || self.usage_snapshot_converged != Some(false)
+            || self.last_update.is_none()
+            || self.scanner_cycle.is_none()
+            || self.scanner_epoch.is_none()
+            || self.usage_snapshot_set_states.is_empty()
+            || u64::try_from(self.buckets_usage.len()).ok() != Some(self.buckets_count)
+        {
+            return false;
+        }
+
+        let mut previous = None;
+        let mut plan_digest = None;
+        let mut has_source = false;
+        for state in &self.usage_snapshot_set_states {
+            if state.scan_plan_digest.is_none()
+                || plan_digest.is_some_and(|digest| Some(digest) != state.scan_plan_digest)
+                || state.scanner_cycle.is_some() != state.scanner_epoch.is_some()
+                || previous.is_some_and(|(pool, set)| (pool, set) >= (state.pool_index, state.set_index))
+            {
+                return false;
+            }
+            previous = Some((state.pool_index, state.set_index));
+            plan_digest = state.scan_plan_digest;
+            has_source |= state.scanner_cycle.is_some() && !state.tombstone;
+        }
+        has_source
+    }
+
     /// Add object metadata to data usage statistics
     pub fn add_object(&mut self, object_path: &str, meta_object: &rustfs_filemeta::MetaObject) {
         // This method is kept for backward compatibility
@@ -2263,6 +2321,55 @@ mod tests {
         assert!(!observed_data_usage_is_newer(&candidate(2, 9, Some(false), true), &authoritative));
         assert!(!observed_data_usage_is_newer(&candidate(2, 11, Some(true), true), &authoritative));
         assert!(!observed_data_usage_is_newer(&candidate(2, 11, Some(false), false), &authoritative));
+
+        let mut partial = candidate(2, 11, Some(false), false);
+        partial.usage_snapshot_partial = true;
+        partial.usage_snapshot_set_states = vec![DataUsageSnapshotSetState {
+            pool_index: 0,
+            set_index: 0,
+            scanner_cycle: Some(10),
+            scanner_epoch: Some(2),
+            scan_plan_digest: Some([1; 32]),
+            complete: false,
+            tombstone: false,
+        }];
+        assert!(observed_data_usage_is_newer(&partial, &authoritative));
+    }
+
+    #[test]
+    fn mixed_topology_snapshot_is_rejected() {
+        let mut partial = DataUsageInfo {
+            last_update: Some(SystemTime::UNIX_EPOCH + Duration::from_secs(2)),
+            scanner_cycle: Some(11),
+            scanner_epoch: Some(2),
+            buckets_count: 0,
+            usage_snapshot_converged: Some(false),
+            usage_snapshot_partial: true,
+            usage_snapshot_set_states: vec![
+                DataUsageSnapshotSetState {
+                    pool_index: 0,
+                    set_index: 0,
+                    scanner_cycle: Some(11),
+                    scanner_epoch: Some(2),
+                    scan_plan_digest: Some([1; 32]),
+                    complete: true,
+                    tombstone: false,
+                },
+                DataUsageSnapshotSetState {
+                    pool_index: 1,
+                    set_index: 0,
+                    scanner_cycle: Some(10),
+                    scanner_epoch: Some(2),
+                    scan_plan_digest: Some([2; 32]),
+                    complete: false,
+                    tombstone: false,
+                },
+            ],
+            ..Default::default()
+        };
+        assert!(!partial.is_valid_partial_snapshot());
+        partial.usage_snapshot_set_states[1].scan_plan_digest = Some([1; 32]);
+        assert!(partial.is_valid_partial_snapshot());
     }
 
     #[test]

crates/e2e_test/src/lib.rs

@@ -61,11 +61,6 @@ mod get_codec_streaming_compat_test;
 #[cfg(test)]
 mod version_id_regression_test;
 
-// Receiver-side replication LWW (rustfs/backlog#1953): stale inbound
-// replication metadata must not overwrite a newer local category state.
-#[cfg(test)]
-mod replication_lww_receiver_test;
-
 // Data usage regression tests
 #[cfg(test)]
 mod data_usage_test;

crates/e2e_test/src/replication_lww_receiver_test.rs

@@ -1,150 +0,0 @@
-#![cfg(test)]
-// 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.
-
-//! Receiver-side replication LWW over the wire (rustfs/backlog#1953, audit
-//! A4/P1-6).
-//!
-//! In an active-active topology both sites' metadata states arrive at the
-//! peer as authorized replication PUTs carrying per-category source
-//! timestamps (`x-rustfs-source-replication-tagging-timestamp` header
-//! family). Before the fix the receiver applied them unconditionally, so a
-//! stale delivery overwrote a newer local state and the two sites diverged
-//! permanently while both reported COMPLETED. This test drives one live
-//! `rustfs` server with simulated inbound replication PUTs for the same
-//! object version and asserts the newer tagging state wins regardless of
-//! delivery order, while a stale delivery still succeeds at the object level
-//! (a failure would loop through MRF re-delivering the stale value).
-
-use crate::common::{RustFSTestEnvironment, init_logging};
-use aws_sdk_s3::Client;
-use aws_sdk_s3::primitives::ByteStream;
-use aws_sdk_s3::types::{BucketVersioningStatus, VersioningConfiguration};
-
-type TestResult = Result<(), Box<dyn std::error::Error + Send + Sync>>;
-
-const HDR_SOURCE_REPLICATION_REQUEST: &str = "x-rustfs-source-replication-request";
-const HDR_SOURCE_VERSION_ID: &str = "x-rustfs-source-version-id";
-const HDR_SOURCE_MTIME: &str = "x-rustfs-source-mtime";
-const HDR_SOURCE_TAGGING_TIMESTAMP: &str = "x-rustfs-source-replication-tagging-timestamp";
-
-const SOURCE_MTIME: &str = "2026-01-01T00:00:00Z";
-const T_STALE: &str = "2026-01-01T00:00:01Z";
-const T_LOCAL: &str = "2026-02-01T00:00:00Z";
-const T_NEWER: &str = "2026-03-01T00:00:00Z";
-
-/// Simulated inbound authorized replication PUT: same object version, tags and
-/// the source-authored tagging timestamp carried in transport headers.
-async fn inbound_replication_put(
-    client: &Client,
-    bucket: &str,
-    key: &str,
-    version_id: &str,
-    tags: &str,
-    tagging_timestamp: &str,
-) -> TestResult {
-    let version_id = version_id.to_string();
-    let tagging_timestamp = tagging_timestamp.to_string();
-    client
-        .put_object()
-        .bucket(bucket)
-        .key(key)
-        .body(ByteStream::from_static(b"lww-e2e-body"))
-        .tagging(tags)
-        .customize()
-        .mutate_request(move |req| {
-            req.headers_mut().insert(HDR_SOURCE_REPLICATION_REQUEST, "true");
-            req.headers_mut().insert(HDR_SOURCE_VERSION_ID, version_id.clone());
-            req.headers_mut().insert(HDR_SOURCE_MTIME, SOURCE_MTIME);
-            req.headers_mut()
-                .insert(HDR_SOURCE_TAGGING_TIMESTAMP, tagging_timestamp.clone());
-        })
-        .send()
-        .await?;
-    Ok(())
-}
-
-async fn tag_value(client: &Client, bucket: &str, key: &str, version_id: &str, tag_key: &str) -> Option<String> {
-    let tagging = client
-        .get_object_tagging()
-        .bucket(bucket)
-        .key(key)
-        .version_id(version_id)
-        .send()
-        .await
-        .expect("object tagging should be readable");
-    tagging
-        .tag_set()
-        .iter()
-        .find(|tag| tag.key() == tag_key)
-        .map(|tag| tag.value().to_string())
-}
-
-#[tokio::test(flavor = "multi_thread")]
-async fn receiver_lww_keeps_newer_tags_across_delivery_orders() -> TestResult {
-    init_logging();
-    let mut env = RustFSTestEnvironment::new().await?;
-    env.start_rustfs_server(vec![]).await?;
-    let client = env.create_s3_client();
-
-    let bucket = "replication-lww-receiver";
-    let key = "object";
-    client.create_bucket().bucket(bucket).send().await?;
-    client
-        .put_bucket_versioning()
-        .bucket(bucket)
-        .versioning_configuration(
-            VersioningConfiguration::builder()
-                .status(BucketVersioningStatus::Enabled)
-                .build(),
-        )
-        .send()
-        .await?;
-
-    // First delivery establishes version V with tags stamped T_LOCAL.
-    let version_id = uuid::Uuid::new_v4().to_string();
-    inbound_replication_put(&client, bucket, key, &version_id, "site=local", T_LOCAL).await?;
-    assert_eq!(
-        tag_value(&client, bucket, key, &version_id, "site").await.as_deref(),
-        Some("local"),
-        "the first delivery must establish the tagged version"
-    );
-
-    // A stale delivery (older source timestamp) must succeed at the object
-    // level but must NOT overwrite the newer tags.
-    inbound_replication_put(&client, bucket, key, &version_id, "site=stale", T_STALE).await?;
-    assert_eq!(
-        tag_value(&client, bucket, key, &version_id, "site").await.as_deref(),
-        Some("local"),
-        "a stale inbound delivery must not overwrite newer tags (rustfs/backlog#1953)"
-    );
-
-    // A newer delivery still converges the version onto the newest state.
-    inbound_replication_put(&client, bucket, key, &version_id, "site=newer", T_NEWER).await?;
-    assert_eq!(
-        tag_value(&client, bucket, key, &version_id, "site").await.as_deref(),
-        Some("newer"),
-        "a newer inbound delivery must overwrite older tags"
-    );
-
-    client
-        .delete_object()
-        .bucket(bucket)
-        .key(key)
-        .version_id(&version_id)
-        .send()
-        .await?;
-    env.delete_test_bucket(bucket).await.ok();
-    Ok(())
-}

crates/ecstore/src/bucket/replication/replication_resyncer.rs

@@ -3868,7 +3868,6 @@ async fn replicate_object_with_multipart<S: ReplicationObjectIO>(ctx: MultipartR
                 actual_size,
                 object_info.etag.clone().unwrap_or_default(),
                 object_info.mod_time,
-                &put_opts.internal,
             ),
         )
         .await

crates/ecstore/src/bucket/replication/replication_target_boundary.rs

@@ -472,7 +472,6 @@ pub(crate) fn replication_complete_multipart_options(
     actual_size: String,
     source_etag: String,
     source_mtime: Option<OffsetDateTime>,
-    source_internal: &AdvancedPutOptions,
 ) -> PutObjectOptions {
     let mut user_metadata = HashMap::new();
     insert_header_map(&mut user_metadata, SUFFIX_REPLICATION_ACTUAL_OBJECT_SIZE, actual_size);
@@ -485,14 +484,6 @@ pub(crate) fn replication_complete_multipart_options(
             // mtime must degrade to epoch so header() suppresses the header
             // instead of asserting the replication time as the object's mtime.
             source_mtime: source_mtime.unwrap_or(OffsetDateTime::UNIX_EPOCH),
-            // Carry the per-category LWW timestamps on the complete request as
-            // well: the receiver's CompleteMultipartUpload options builder
-            // parses the same headers, so the multipart transport gets the
-            // same receiver-side LWW as the single-PUT transport
-            // (rustfs/backlog#1953). Epoch values keep the headers suppressed.
-            tagging_timestamp: source_internal.tagging_timestamp,
-            retention_timestamp: source_internal.retention_timestamp,
-            legalhold_timestamp: source_internal.legalhold_timestamp,
             replication_status: ReplicationStatusType::Replica,
             replication_request: true,
             ..Default::default()
@@ -672,39 +663,20 @@ mod tests {
     #[test]
     fn replication_complete_multipart_options_sets_actual_size() {
         let source_mtime = OffsetDateTime::from_unix_timestamp(1_716_170_000).expect("valid test timestamp");
-        let source_internal = AdvancedPutOptions {
-            tagging_timestamp: OffsetDateTime::from_unix_timestamp(1_716_170_100).expect("valid test timestamp"),
-            retention_timestamp: OffsetDateTime::from_unix_timestamp(1_716_170_200).expect("valid test timestamp"),
-            legalhold_timestamp: OffsetDateTime::from_unix_timestamp(1_716_170_300).expect("valid test timestamp"),
-            ..Default::default()
-        };
         let options = replication_complete_multipart_options(
             "1024".to_string(),
             "0123456789abcdef0123456789abcdef-3".to_string(),
             Some(source_mtime),
-            &source_internal,
         );
         assert_eq!(options.internal.source_etag, "0123456789abcdef0123456789abcdef-3");
         assert_eq!(options.internal.source_mtime, source_mtime);
 
-        // The complete request must carry the same per-category LWW timestamps
-        // as the initiate request; the receiver reads them from the complete
-        // headers (rustfs/backlog#1953).
-        assert_eq!(options.internal.tagging_timestamp, source_internal.tagging_timestamp);
-        assert_eq!(options.internal.retention_timestamp, source_internal.retention_timestamp);
-        assert_eq!(options.internal.legalhold_timestamp, source_internal.legalhold_timestamp);
-
         // Absent source mtime must degrade to epoch (header suppressed), not
         // the AdvancedPutOptions default of now_utc() — that default would
         // stamp the replication time as the replica's mtime and break the
-        // multipart HEAD convergence. Unset category timestamps stay epoch so
-        // header() keeps suppressing them.
-        let options_no_mtime =
-            replication_complete_multipart_options("1024".to_string(), String::new(), None, &AdvancedPutOptions::default());
+        // multipart HEAD convergence.
+        let options_no_mtime = replication_complete_multipart_options("1024".to_string(), String::new(), None);
         assert_eq!(options_no_mtime.internal.source_mtime.unix_timestamp(), 0);
-        assert_eq!(options_no_mtime.internal.tagging_timestamp.unix_timestamp(), 0);
-        assert_eq!(options_no_mtime.internal.retention_timestamp.unix_timestamp(), 0);
-        assert_eq!(options_no_mtime.internal.legalhold_timestamp.unix_timestamp(), 0);
 
         assert_eq!(
             get_header_map(&options.user_metadata, SUFFIX_REPLICATION_ACTUAL_OBJECT_SIZE).as_deref(),

crates/ecstore/src/data_usage/mod.rs

@@ -73,6 +73,16 @@ struct CachedBucketUsage {
     // mutation. A strictly later generation is required before the mutation
     // evidence can be discarded.
     pending_scanner_position: Option<(u64, u64)>,
+    // Deletes are visible to admin immediately, but quota admission keeps
+    // them pending until a complete scanner generation reconciles the set.
+    // This marker intentionally remains process-local: the delete request
+    // updates this overlay before the scanner writes a durable snapshot. If
+    // the process restarts first, loading the persisted complete snapshot
+    // restores the pre-reconciliation (larger) baseline, which is
+    // conservative for quota admission. A persisted post-delete snapshot is
+    // necessarily a complete scanner reconciliation and therefore creates a
+    // fresh cache entry with no pending hold.
+    pending_negative_delta: u64,
 }
 
 type UsageMemoryCache = Arc<RwLock<HashMap<String, CachedBucketUsage>>>;
@@ -948,7 +958,12 @@ async fn load_observed_data_usage_snapshot(store: Arc<ECStore>) -> Option<DataUs
     };
 
     match parse_usage_snapshot(&data) {
-        Ok(info) if info.usage_snapshot_converged == Some(false) && info.is_complete_bucket_usage_snapshot() => Some(info),
+        Ok(info)
+            if info.usage_snapshot_converged == Some(false)
+                && (info.is_complete_bucket_usage_snapshot() || info.is_valid_partial_snapshot()) =>
+        {
+            Some(info)
+        }
         Ok(_) => {
             error!(
                 event = "data_usage_snapshot_load_failed",
@@ -993,7 +1008,7 @@ async fn load_admin_data_usage_from_backend(store: Arc<ECStore>) -> Result<DataU
 }
 
 fn discard_incomplete_bucket_usage(data_usage_info: &mut DataUsageInfo) {
-    if !data_usage_info.is_complete_bucket_usage_snapshot() {
+    if !data_usage_info.is_complete_bucket_usage_snapshot() && !data_usage_info.usage_snapshot_partial {
         data_usage_info.usage_snapshot_complete = false;
         data_usage_info.buckets_usage.clear();
         data_usage_info.bucket_sizes.clear();
@@ -1643,6 +1658,7 @@ fn cached_bucket_usage_from_backend(usage: BucketUsageInfo, updated_at: SystemTi
         dirty: false,
         stale_snapshot_pending: false,
         pending_scanner_position: None,
+        pending_negative_delta: 0,
     }
 }
 
@@ -1656,6 +1672,7 @@ fn cached_bucket_usage_now(usage: BucketUsageInfo) -> CachedBucketUsage {
         dirty: false,
         stale_snapshot_pending: false,
         pending_scanner_position: None,
+        pending_negative_delta: 0,
     }
 }
 
@@ -1808,6 +1825,7 @@ pub async fn record_bucket_object_delete_memory(bucket: &str, deleted_size: u64,
         .or_insert_with(|| cached_bucket_usage_now(BucketUsageInfo::default()));
 
     entry.usage.size = entry.usage.size.saturating_sub(deleted_size);
+    entry.pending_negative_delta = entry.pending_negative_delta.saturating_add(deleted_size);
     if removed_current_object {
         entry.usage.objects_count = entry.usage.objects_count.saturating_sub(1);
         entry.usage.versions_count = entry.usage.versions_count.saturating_sub(1);
@@ -1863,7 +1881,7 @@ pub async fn get_bucket_usage_memory(bucket: &str) -> Option<u64> {
     cache
         .get(bucket)
         .filter(|cached| cached.authoritative)
-        .map(|cached| cached.usage.size)
+        .map(|cached| cached.usage.size.saturating_add(cached.pending_negative_delta))
 }
 
 async fn update_usage_cache_if_needed() {
@@ -2943,6 +2961,45 @@ mod tests {
         assert_eq!(selected.usage_snapshot_converged, Some(true));
     }
 
+    #[test]
+    fn persisted_authoritative_stalls_but_memory_overlay_remains_visible() {
+        let authoritative = DataUsageInfo {
+            last_update: Some(SystemTime::UNIX_EPOCH),
+            scanner_epoch: Some(4),
+            scanner_cycle: Some(10),
+            usage_snapshot_complete: true,
+            ..Default::default()
+        };
+        let mut partial = authoritative.clone();
+        partial.last_update = Some(SystemTime::UNIX_EPOCH + Duration::from_secs(1));
+        partial.scanner_cycle = Some(11);
+        partial.usage_snapshot_complete = false;
+        partial.usage_snapshot_partial = true;
+        partial.usage_snapshot_converged = Some(false);
+        partial.usage_snapshot_authoritative_baseline = Some(authoritative.snapshot_identity());
+        partial.usage_snapshot_set_states = vec![rustfs_data_usage::DataUsageSnapshotSetState {
+            pool_index: 0,
+            set_index: 0,
+            scanner_cycle: Some(10),
+            scanner_epoch: Some(4),
+            scan_plan_digest: Some([1; 32]),
+            complete: false,
+            tombstone: false,
+        }];
+        partial.buckets_usage.insert(
+            "bucket".to_string(),
+            BucketUsageInfo {
+                size: 100,
+                ..Default::default()
+            },
+        );
+        partial.buckets_count = 1;
+
+        let (selected, _) = select_admin_data_usage_snapshot(authoritative, true, Some(partial));
+        assert!(selected.usage_snapshot_partial);
+        assert_eq!(selected.buckets_usage.get("bucket").map(|usage| usage.size), Some(100));
+    }
+
     #[tokio::test]
     async fn authoritative_save_cleanup_removes_observed_snapshot_best_effort() {
         let store = UsageCasStore::default();
@@ -4665,6 +4722,55 @@ mod tests {
         );
     }
 
+    #[tokio::test]
+    #[serial]
+    async fn partial_usage_is_observational_not_authoritative_for_quota() {
+        clear_usage_memory_cache_for_test().await;
+
+        let mut partial = data_usage_info_for_test("bucket-a", 10, 100, SystemTime::now());
+        partial.usage_snapshot_complete = false;
+        partial.usage_snapshot_partial = true;
+        replace_bucket_usage_memory_from_info(&partial).await;
+
+        assert_eq!(get_bucket_usage_memory("bucket-a").await, None);
+    }
+
+    #[tokio::test]
+    #[serial]
+    async fn stale_quota_uses_complete_baseline_plus_positive_deltas() {
+        clear_usage_memory_cache_for_test().await;
+
+        let baseline = data_usage_info_for_test("bucket-a", 1, 100, SystemTime::now());
+        replace_bucket_usage_memory_from_info(&baseline).await;
+        record_bucket_object_write_memory("bucket-a", None, 25).await;
+
+        assert_eq!(get_bucket_usage_memory("bucket-a").await, Some(125));
+    }
+
+    #[tokio::test]
+    #[serial]
+    async fn negative_delta_waits_for_set_reconciliation() {
+        clear_usage_memory_cache_for_test().await;
+
+        let baseline = data_usage_info_for_test("bucket-a", 1, 100, SystemTime::UNIX_EPOCH + Duration::from_secs(100));
+        replace_bucket_usage_memory_from_info(&baseline).await;
+        record_bucket_object_delete_memory("bucket-a", 25, true).await;
+
+        assert_eq!(get_bucket_usage_memory("bucket-a").await, Some(100));
+
+        // Simulate a process restart: the request-path overlay is gone, but
+        // the persisted authoritative snapshot is still the pre-reconciliation
+        // baseline. Quota must remain conservative until a complete scanner
+        // result proves the delete.
+        clear_usage_memory_cache_for_test().await;
+        replace_bucket_usage_memory_from_info(&baseline).await;
+        assert_eq!(get_bucket_usage_memory("bucket-a").await, Some(100));
+
+        let reconciled = data_usage_info_for_test("bucket-a", 0, 75, SystemTime::UNIX_EPOCH + Duration::from_secs(101));
+        replace_bucket_usage_memory_from_info(&reconciled).await;
+        assert_eq!(get_bucket_usage_memory("bucket-a").await, Some(75));
+    }
+
     #[tokio::test]
     #[serial]
     async fn memory_overlay_counts_versioned_overwrite_as_new_version() {

crates/ecstore/src/set_disk/ops/multipart.rs

@@ -2318,57 +2318,6 @@ impl crate::storage_api_contracts::multipart::MultipartOperations for SetDisks {
             fi.set_data_moved();
         }
 
-        // Receiver-side LWW (rustfs/backlog#1953): the multipart replication
-        // transport carries the category values at CreateMultipartUpload (in
-        // the staged upload metadata) and the source category timestamps on
-        // the complete request. Read the destination version under the held
-        // object write lock and keep any category this site modified more
-        // recently. Read failures (version absent on first replication, quorum
-        // errors) keep today's overwrite semantics: failing the complete would
-        // loop through MRF, re-delivering the stale value forever.
-        if crate::set_disk::ops::object::replication_lww_applicable(opts)
-            && let Some(version_id) = fi.version_id
-        {
-            match self
-                .get_object_info(
-                    bucket,
-                    object,
-                    &ObjectOptions {
-                        version_id: Some(version_id.to_string()),
-                        no_lock: true,
-                        metadata_cache_safe: false,
-                        versioned: opts.versioned,
-                        version_suspended: opts.version_suspended,
-                        ..Default::default()
-                    },
-                )
-                .await
-            {
-                Ok(existing) => {
-                    let stored = crate::set_disk::ops::object::stored_replication_category_metadata(&existing);
-                    crate::set_disk::ops::object::merge_replication_metadata_lww(&mut fi.metadata, &stored, opts);
-                }
-                // Version absent: first replication of this version, nothing
-                // local to compare — the normal path, not a degraded one.
-                Err(err) if is_err_object_not_found(&err) || is_err_version_not_found(&err) => {}
-                Err(err) => {
-                    // Degraded path: without the stored state the inbound
-                    // metadata is applied unchanged — exactly the overwrite
-                    // LWW exists to prevent — so this must be operator-visible.
-                    warn!(
-                        component = LOG_COMPONENT_ECSTORE,
-                        subsystem = LOG_SUBSYSTEM_SET_DISK,
-                        bucket,
-                        object,
-                        version_id = %version_id,
-                        error = %err,
-                        state = "replication_lww_read_unavailable",
-                        "SetDisk multipart replication LWW read skipped; inbound metadata applied without comparison"
-                    );
-                }
-            }
-        }
-
         for meta in parts_metadatas.iter_mut() {
             if meta.has_valid_erasure_geometry() {
                 meta.size = fi.size;
@@ -7106,97 +7055,6 @@ mod tests {
                 .await
         }
 
-        /// Receiver-side LWW on the multipart replication transport
-        /// (rustfs/backlog#1953): a metadata-only replication of a multipart
-        /// source object rides CreateMultipartUpload (category values in the
-        /// upload metadata) + CompleteMultipartUpload (category timestamps in
-        /// the complete options). A stale inbound tagging timestamp must not
-        /// overwrite a newer locally-tagged destination version.
-        #[tokio::test]
-        #[serial]
-        async fn complete_multipart_upload_stale_replication_tags_keep_local() {
-            use rustfs_utils::http::headers::AMZ_OBJECT_TAGGING;
-            use rustfs_utils::http::{SUFFIX_TAGGING_TIMESTAMP, get_str};
-            use time::format_description::well_known::Rfc3339;
-
-            const T_OLD: &str = "2026-01-01T00:00:00Z";
-            const T_LOCAL: &str = "2026-02-01T00:00:00Z";
-
-            let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-            let bucket = "multipart-replication-lww-bucket";
-            let object = "object";
-            make_bucket_on_all(&disk_stores, bucket).await;
-
-            // Local destination version with newer tags.
-            let version_id = Uuid::new_v4();
-            let mut local_metadata = HashMap::new();
-            local_metadata.insert(AMZ_OBJECT_TAGGING.to_string(), "site=local".to_string());
-            rustfs_utils::http::insert_str(&mut local_metadata, SUFFIX_TAGGING_TIMESTAMP, T_LOCAL.to_string());
-            let mut local_reader = PutObjReader::from_vec(b"local body".to_vec());
-            set_disks
-                .put_object(
-                    bucket,
-                    object,
-                    &mut local_reader,
-                    &ObjectOptions {
-                        versioned: true,
-                        version_id: Some(version_id.to_string()),
-                        user_defined: local_metadata,
-                        // Explicit-version PUTs require the bucket Object Lock snapshot.
-                        object_lock_config_snapshot: Some(Arc::new(crate::set_disk::ObjectLockConfigSnapshot::new(
-                            crate::bucket::metadata_sys::ObjectLockConfigState::ConfirmedAbsent,
-                        ))),
-                        ..Default::default()
-                    },
-                )
-                .await
-                .expect("local versioned put should commit");
-
-            // Inbound replication upload carrying older tags for the same version.
-            let mut inbound_metadata = HashMap::new();
-            inbound_metadata.insert(AMZ_OBJECT_TAGGING.to_string(), "site=remote".to_string());
-            rustfs_utils::http::insert_str(&mut inbound_metadata, SUFFIX_TAGGING_TIMESTAMP, T_OLD.to_string());
-            let create_opts = ObjectOptions {
-                versioned: true,
-                user_defined: inbound_metadata,
-                ..Default::default()
-            };
-            let (upload_id, parts) =
-                stage_upload_with_create_opts(&set_disks, bucket, object, &payload(0x5a), &create_opts).await;
-            rewrite_staged_upload_version_id(&set_disks, bucket, object, &upload_id, Some(version_id)).await;
-
-            let complete_opts = ObjectOptions {
-                versioned: true,
-                replication_request: true,
-                replication_tagging_timestamp: Some(OffsetDateTime::parse(T_OLD, &Rfc3339).expect("test timestamp should parse")),
-                ..Default::default()
-            };
-            set_disks
-                .clone()
-                .complete_multipart_upload(bucket, object, &upload_id, parts, &complete_opts)
-                .await
-                .expect("replication multipart completion should succeed even when a category keeps local values");
-
-            let info = set_disks
-                .get_object_info(
-                    bucket,
-                    object,
-                    &ObjectOptions {
-                        versioned: true,
-                        version_id: Some(version_id.to_string()),
-                        ..Default::default()
-                    },
-                )
-                .await
-                .expect("completed version should be readable");
-            assert_eq!(
-                info.user_tags.as_str(),
-                "site=local",
-                "older inbound multipart tags must not overwrite newer local tags"
-            );
-            assert_eq!(get_str(&info.user_defined, SUFFIX_TAGGING_TIMESTAMP).as_deref(), Some(T_LOCAL));
-        }
-
         #[tokio::test]
         #[serial]
         async fn complete_multipart_upload_assigns_completion_version_id() {

crates/ecstore/src/set_disk/ops/object.rs

@@ -1881,110 +1881,6 @@ fn delete_file_info_with_replication_transport_metadata(fi: &FileInfo) -> FileIn
     transported
 }
 
-/// True when an authorized replication write carries at least one per-category
-/// source timestamp, i.e. receiver-side LWW has something to judge.
-pub(in crate::set_disk) fn replication_lww_applicable(opts: &ObjectOptions) -> bool {
-    opts.replication_request
-        && (opts.replication_tagging_timestamp.is_some()
-            || opts.replication_retention_timestamp.is_some()
-            || opts.replication_legalhold_timestamp.is_some())
-}
-
-/// The stored per-category state of a destination version, as compared by
-/// [`merge_replication_metadata_lww`]. `ObjectInfo::from_file_info`
-/// externalizes tags into `user_tags` (stripping the metadata key), so the
-/// tag value is folded back into map form here.
-pub(in crate::set_disk) fn stored_replication_category_metadata(existing: &ObjectInfo) -> HashMap<String, String> {
-    let mut stored = (*existing.user_defined).clone();
-    if !existing.user_tags.is_empty() {
-        stored.insert(rustfs_utils::http::headers::AMZ_OBJECT_TAGGING.to_string(), (*existing.user_tags).clone());
-    }
-    stored
-}
-
-/// Receiver-side last-writer-wins for authorized replication writes
-/// (rustfs/backlog#1953, audit A4/P1-6). Metadata-only replication reuses the
-/// whole-object transports, so in active-active topologies an inbound write
-/// carries the source's tags / retention / legal hold verbatim and would
-/// otherwise overwrite a category the destination modified more recently —
-/// both sites end up permanently diverged while reporting COMPLETED.
-///
-/// Judged per category, only when the inbound request carries that category's
-/// source timestamp (`ObjectOptions::replication_*_timestamp`):
-/// - stored timestamp newer than inbound: the local category values and
-///   timestamp are kept; the rest of the write proceeds per the inbound
-///   metadata and the object-level result stays successful (failing the write
-///   instead would loop through MRF, re-delivering the stale value forever);
-/// - otherwise the inbound category wins and its internal timestamp key is
-///   pinned to the source-authored time — the PUT path re-stamps the
-///   object-lock timestamps with the receiver's clock
-///   (`parse_object_lock_retention` / `parse_object_lock_legal_hold` insert
-///   `now()` via `eval_metadata`), which would make the replica's clock the
-///   LWW authority and wedge later convergence;
-/// - no stored timestamp (pre-P1-6 data) or no inbound timestamp: the current
-///   overwrite behavior is preserved.
-///
-/// Returns whether `inbound` was modified. Callers must hold the object write
-/// lock so the stored values compared here are the ones being replaced.
-pub(in crate::set_disk) fn merge_replication_metadata_lww(
-    inbound: &mut HashMap<String, String>,
-    existing: &HashMap<String, String>,
-    opts: &ObjectOptions,
-) -> bool {
-    use rustfs_utils::http::headers::{
-        AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER, AMZ_OBJECT_LOCK_MODE_LOWER, AMZ_OBJECT_LOCK_RETAIN_UNTIL_DATE_LOWER, AMZ_OBJECT_TAGGING,
-    };
-    use rustfs_utils::http::metadata_compat::{
-        SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, SUFFIX_TAGGING_TIMESTAMP, get_str,
-        remove_str,
-    };
-    use time::format_description::well_known::Rfc3339;
-
-    let categories: [(Option<OffsetDateTime>, &str, &[&str]); 3] = [
-        (opts.replication_tagging_timestamp, SUFFIX_TAGGING_TIMESTAMP, &[AMZ_OBJECT_TAGGING]),
-        (
-            opts.replication_retention_timestamp,
-            SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP,
-            &[AMZ_OBJECT_LOCK_MODE_LOWER, AMZ_OBJECT_LOCK_RETAIN_UNTIL_DATE_LOWER],
-        ),
-        (
-            opts.replication_legalhold_timestamp,
-            SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP,
-            &[AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER],
-        ),
-    ];
-
-    let mut changed = false;
-    for (inbound_timestamp, timestamp_suffix, value_keys) in categories {
-        let Some(inbound_timestamp) = inbound_timestamp else { continue };
-        let is_category_value_key = |key: &str| value_keys.iter().any(|value_key| key.eq_ignore_ascii_case(value_key));
-        let stored_timestamp = get_str(existing, timestamp_suffix).and_then(|value| OffsetDateTime::parse(&value, &Rfc3339).ok());
-        if stored_timestamp.is_some_and(|stored| stored > inbound_timestamp) {
-            inbound.retain(|key, _| !is_category_value_key(key));
-            remove_str(inbound, timestamp_suffix);
-            for (key, value) in existing {
-                if is_category_value_key(key) {
-                    inbound.insert(key.clone(), value.clone());
-                }
-            }
-            // Restore the winning timestamp via insert_str, not a verbatim key
-            // copy: a MinIO-written version may carry only the
-            // x-minio-internal- key, and the dual-key invariant requires every
-            // write to produce both keys.
-            if let Some(stored_value) = get_str(existing, timestamp_suffix) {
-                rustfs_utils::http::insert_str(inbound, timestamp_suffix, stored_value);
-            }
-            changed = true;
-        } else if let Ok(source_authored) = inbound_timestamp.format(&Rfc3339)
-            && get_str(inbound, timestamp_suffix).as_deref() != Some(source_authored.as_str())
-        {
-            rustfs_utils::http::insert_str(inbound, timestamp_suffix, source_authored);
-            changed = true;
-        }
-    }
-    changed
-}
-
 impl SetDisks {
     pub(in crate::set_disk) async fn persist_old_data_cleanup_receipts(
         &self,
@@ -2228,13 +2124,26 @@ impl SetDisks {
 
             let put_object_size = known_put_object_storage_size(data.size());
             let shard_file_size_raw = erasure.shard_file_size(put_object_size);
-            let is_inline_buffer = storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);
+            let is_inline_buffer =
+                storage_class_config.should_inline(shard_file_size_raw, erasure.data_shards, opts.versioned);
 
             let collect_stage_timing = rustfs_io_metrics::put_stage_metrics_enabled() || issue3031_diag_enabled();
             let shard_file_size = shard_file_size_raw;
             let shard_size = erasure.shard_size();
             let write_path = classify_put_write_path(is_inline_buffer, put_object_size, fi.erasure.block_size);
             let direct_inline_commit = matches!(write_path, SmallWritePath::Inline);
+            {
+                use std::io::Write;
+                let msg = format!(
+                    "INLINE_DEBUG: bucket={} obj={} size={} shard_fs={} ds={} bs={} inline={} direct={} path={} iblock={} ver={}\n",
+                    bucket, object, put_object_size, shard_file_size_raw, erasure.data_shards, fi.erasure.block_size,
+                    is_inline_buffer, direct_inline_commit, write_path.metric_label(), storage_class_config.inline_block(), opts.versioned
+                );
+                if let Ok(mut f) = std::fs::OpenOptions::new().create(true).append(true).open("/tmp/rustfs_inline_debug.log") {
+                    let _ = f.write_all(msg.as_bytes());
+                }
+                let _ = std::io::stderr().write_all(msg.as_bytes());
+            }
             rustfs_io_metrics::record_put_object_path(write_path.metric_label());
             let writer_setup_stage_start = collect_stage_timing.then(Instant::now);
             let (mut writers, errors) = if direct_inline_commit {
@@ -2666,22 +2575,6 @@ impl SetDisks {
                         if check_object_lock_for_deletion_with_state(object_lock_config.state(), &existing, false)?.is_some() {
                             return Err(StorageError::PrefixAccessDenied(bucket.to_string(), object.to_string()));
                         }
-                        // Receiver-side LWW (rustfs/backlog#1953): reuse this
-                        // commit-lock read of the destination version so a
-                        // category (tags / retention / legal hold) modified
-                        // more recently on this site is kept instead of being
-                        // overwritten by the inbound replication metadata.
-                        if replication_lww_applicable(opts) {
-                            let stored = stored_replication_category_metadata(&existing);
-                            let mut merged = parts_metadatas[response_metadata_slot].metadata.clone();
-                            if merge_replication_metadata_lww(&mut merged, &stored, opts) {
-                                for (pfi, disk) in parts_metadatas.iter_mut().zip(shuffle_disks.iter()) {
-                                    if disk.is_some() {
-                                        pfi.metadata = merged.clone();
-                                    }
-                                }
-                            }
-                        }
                     }
                     Err(err) if is_err_object_not_found(&err) || is_err_version_not_found(&err) => {}
                     Err(err) => return Err(err),
@@ -8186,357 +8079,6 @@ mod replication_quota_safety_tests {
     }
 }
 
-#[cfg(test)]
-mod replication_lww_tests {
-    //! Receiver-side LWW for authorized replication writes (rustfs/backlog#1953,
-    //! audit A4/P1-6): an inbound replication PUT whose per-category timestamp
-    //! (tags / retention / legal hold) is older than the destination version's
-    //! stored timestamp must keep the local category values instead of
-    //! overwriting them; categories are judged independently and the write
-    //! itself still succeeds.
-
-    use super::hermetic_set_disks_support::hermetic_set_disks_isolated as hermetic_set_disks;
-    use super::*;
-    use crate::storage_api_contracts::object::{ObjectIO as _, ObjectOperations as _};
-    use rustfs_utils::http::headers::{
-        AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER, AMZ_OBJECT_LOCK_MODE_LOWER, AMZ_OBJECT_LOCK_RETAIN_UNTIL_DATE_LOWER, AMZ_OBJECT_TAGGING,
-    };
-    use rustfs_utils::http::{
-        SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, SUFFIX_TAGGING_TIMESTAMP, get_str,
-        insert_str,
-    };
-    use time::format_description::well_known::Rfc3339;
-
-    const T_OLD: &str = "2026-01-01T00:00:00Z";
-    const T_LOCAL: &str = "2026-02-01T00:00:00Z";
-    const T_NEW: &str = "2026-03-01T00:00:00Z";
-
-    fn parse_ts(value: &str) -> OffsetDateTime {
-        OffsetDateTime::parse(value, &Rfc3339).expect("test timestamp should parse")
-    }
-
-    async fn make_bucket(disks: &[DiskStore], bucket: &str) {
-        for disk in disks {
-            disk.make_volume(bucket).await.expect("bucket volume should be created");
-        }
-    }
-
-    async fn put_version(set_disks: &Arc<SetDisks>, bucket: &str, object: &str, version_id: &str, opts: &ObjectOptions) {
-        let mut reader = PutObjReader::from_vec(b"lww-body".to_vec());
-        set_disks
-            .put_object(bucket, object, &mut reader, opts)
-            .await
-            .expect("versioned put should commit");
-        assert_eq!(opts.version_id.as_deref(), Some(version_id));
-    }
-
-    fn versioned_opts(version_id: &str, user_defined: HashMap<String, String>) -> ObjectOptions {
-        ObjectOptions {
-            versioned: true,
-            version_id: Some(version_id.to_string()),
-            user_defined,
-            // Explicit-version PUTs require the bucket Object Lock snapshot.
-            object_lock_config_snapshot: Some(Arc::new(ObjectLockConfigSnapshot::new(
-                crate::bucket::metadata_sys::ObjectLockConfigState::ConfirmedAbsent,
-            ))),
-            ..Default::default()
-        }
-    }
-
-    /// Local state: version `version_id` with tags "site=local" stamped `T_LOCAL`.
-    async fn seed_local_tagged_version(set_disks: &Arc<SetDisks>, bucket: &str, object: &str, version_id: &str) {
-        let mut user_defined = HashMap::new();
-        user_defined.insert(AMZ_OBJECT_TAGGING.to_string(), "site=local".to_string());
-        insert_str(&mut user_defined, SUFFIX_TAGGING_TIMESTAMP, T_LOCAL.to_string());
-        put_version(set_disks, bucket, object, version_id, &versioned_opts(version_id, user_defined)).await;
-    }
-
-    fn inbound_tagging_opts(version_id: &str, tags: &str, timestamp: &str) -> ObjectOptions {
-        let mut user_defined = HashMap::new();
-        user_defined.insert(AMZ_OBJECT_TAGGING.to_string(), tags.to_string());
-        insert_str(&mut user_defined, SUFFIX_TAGGING_TIMESTAMP, timestamp.to_string());
-        ObjectOptions {
-            replication_request: true,
-            replication_tagging_timestamp: Some(parse_ts(timestamp)),
-            ..versioned_opts(version_id, user_defined)
-        }
-    }
-
-    async fn version_info(set_disks: &Arc<SetDisks>, bucket: &str, object: &str, version_id: &str) -> ObjectInfo {
-        set_disks
-            .get_object_info(bucket, object, &versioned_opts(version_id, HashMap::new()))
-            .await
-            .expect("version should be readable")
-    }
-
-    #[tokio::test]
-    async fn inbound_stale_tagging_keeps_newer_local_tags() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-tagging-stale";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-        seed_local_tagged_version(&set_disks, bucket, object, &version_id).await;
-
-        put_version(
-            &set_disks,
-            bucket,
-            object,
-            &version_id,
-            &inbound_tagging_opts(&version_id, "site=remote", T_OLD),
-        )
-        .await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(
-            info.user_tags.as_str(),
-            "site=local",
-            "older inbound tags must not overwrite newer local tags"
-        );
-        assert_eq!(
-            get_str(&info.user_defined, SUFFIX_TAGGING_TIMESTAMP).as_deref(),
-            Some(T_LOCAL),
-            "the winning local tagging timestamp must be preserved"
-        );
-    }
-
-    #[tokio::test]
-    async fn inbound_newer_tagging_overwrites_local_tags() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-tagging-newer";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-        seed_local_tagged_version(&set_disks, bucket, object, &version_id).await;
-
-        put_version(
-            &set_disks,
-            bucket,
-            object,
-            &version_id,
-            &inbound_tagging_opts(&version_id, "site=remote", T_NEW),
-        )
-        .await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(
-            info.user_tags.as_str(),
-            "site=remote",
-            "newer inbound tags must overwrite older local tags"
-        );
-        assert_eq!(get_str(&info.user_defined, SUFFIX_TAGGING_TIMESTAMP).as_deref(), Some(T_NEW));
-    }
-
-    #[tokio::test]
-    async fn inbound_wins_when_local_has_no_tagging_timestamp() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-tagging-no-local-ts";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-        // Pre-P1-6 data: local tags without a stored tagging timestamp.
-        let mut user_defined = HashMap::new();
-        user_defined.insert(AMZ_OBJECT_TAGGING.to_string(), "site=local".to_string());
-        put_version(&set_disks, bucket, object, &version_id, &versioned_opts(&version_id, user_defined)).await;
-
-        put_version(
-            &set_disks,
-            bucket,
-            object,
-            &version_id,
-            &inbound_tagging_opts(&version_id, "site=remote", T_OLD),
-        )
-        .await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(
-            info.user_tags.as_str(),
-            "site=remote",
-            "without a local timestamp the inbound category must win (pre-LWW data compatibility)"
-        );
-    }
-
-    #[tokio::test]
-    async fn categories_are_judged_independently() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-category-independent";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-
-        // Local: newer tags (T_LOCAL), older *cleared* retention (T_OLD) —
-        // timestamp key only, the shape a replicated retention clear stores.
-        // (An active local retention would already block the overwrite at the
-        // WORM gate; the LWW-reachable retention states are cleared/expired.)
-        let mut local = HashMap::new();
-        local.insert(AMZ_OBJECT_TAGGING.to_string(), "site=local".to_string());
-        insert_str(&mut local, SUFFIX_TAGGING_TIMESTAMP, T_LOCAL.to_string());
-        insert_str(&mut local, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, T_OLD.to_string());
-        put_version(&set_disks, bucket, object, &version_id, &versioned_opts(&version_id, local)).await;
-
-        // Inbound: older tags (T_OLD), newer retention (T_NEW).
-        let mut inbound = HashMap::new();
-        inbound.insert(AMZ_OBJECT_TAGGING.to_string(), "site=remote".to_string());
-        insert_str(&mut inbound, SUFFIX_TAGGING_TIMESTAMP, T_OLD.to_string());
-        inbound.insert(AMZ_OBJECT_LOCK_MODE_LOWER.to_string(), "COMPLIANCE".to_string());
-        inbound.insert(AMZ_OBJECT_LOCK_RETAIN_UNTIL_DATE_LOWER.to_string(), "2028-01-01T00:00:00Z".to_string());
-        insert_str(&mut inbound, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, T_NEW.to_string());
-        let opts = ObjectOptions {
-            replication_request: true,
-            replication_tagging_timestamp: Some(parse_ts(T_OLD)),
-            replication_retention_timestamp: Some(parse_ts(T_NEW)),
-            ..versioned_opts(&version_id, inbound)
-        };
-        put_version(&set_disks, bucket, object, &version_id, &opts).await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(info.user_tags.as_str(), "site=local", "the stale tagging category must keep local values");
-        assert_eq!(
-            info.user_defined.get(AMZ_OBJECT_LOCK_MODE_LOWER).map(String::as_str),
-            Some("COMPLIANCE"),
-            "the newer retention category must be applied in the same write"
-        );
-        assert_eq!(get_str(&info.user_defined, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP).as_deref(), Some(T_NEW));
-    }
-
-    #[tokio::test]
-    async fn inbound_stale_legal_hold_keeps_local_value() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-legalhold-stale";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-
-        // Local: legal hold released (OFF) at T_LOCAL. (A local hold that is
-        // still ON already blocks the overwrite at the WORM gate; the
-        // LWW-reachable divergence is a stale inbound ON resurrecting a hold
-        // that was released more recently on this site.)
-        let mut local = HashMap::new();
-        local.insert(AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER.to_string(), "OFF".to_string());
-        insert_str(&mut local, SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP, T_LOCAL.to_string());
-        put_version(&set_disks, bucket, object, &version_id, &versioned_opts(&version_id, local)).await;
-
-        let mut inbound = HashMap::new();
-        inbound.insert(AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER.to_string(), "ON".to_string());
-        insert_str(&mut inbound, SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP, T_OLD.to_string());
-        let opts = ObjectOptions {
-            replication_request: true,
-            replication_legalhold_timestamp: Some(parse_ts(T_OLD)),
-            ..versioned_opts(&version_id, inbound)
-        };
-        put_version(&set_disks, bucket, object, &version_id, &opts).await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(
-            info.user_defined.get(AMZ_OBJECT_LOCK_LEGAL_HOLD_LOWER).map(String::as_str),
-            Some("OFF"),
-            "a stale inbound legal hold must not resurrect a hold released more recently"
-        );
-        assert_eq!(
-            get_str(&info.user_defined, SUFFIX_OBJECTLOCK_LEGALHOLD_TIMESTAMP).as_deref(),
-            Some(T_LOCAL)
-        );
-    }
-
-    /// Dual-key invariant under LWW: a MinIO-written destination version may
-    /// carry only the x-minio-internal timestamp key; when the local category
-    /// wins, the restored map must still hold BOTH compatibility keys.
-    #[test]
-    fn local_win_restores_both_internal_timestamp_keys_for_minio_only_metadata() {
-        let mut inbound = HashMap::new();
-        inbound.insert(AMZ_OBJECT_TAGGING.to_string(), "site=remote".to_string());
-        insert_str(&mut inbound, SUFFIX_TAGGING_TIMESTAMP, T_OLD.to_string());
-        let existing = HashMap::from([
-            (AMZ_OBJECT_TAGGING.to_string(), "site=local".to_string()),
-            ("X-Minio-Internal-Tagging-Timestamp".to_string(), T_LOCAL.to_string()),
-        ]);
-        let opts = ObjectOptions {
-            replication_request: true,
-            replication_tagging_timestamp: Some(parse_ts(T_OLD)),
-            ..Default::default()
-        };
-
-        assert!(merge_replication_metadata_lww(&mut inbound, &existing, &opts));
-        assert_eq!(inbound.get(AMZ_OBJECT_TAGGING).map(String::as_str), Some("site=local"));
-        assert_eq!(
-            inbound.get("x-rustfs-internal-tagging-timestamp").map(String::as_str),
-            Some(T_LOCAL),
-            "the RustFS twin key must be materialized even when the source version only had the MinIO key"
-        );
-        assert_eq!(inbound.get("x-minio-internal-tagging-timestamp").map(String::as_str), Some(T_LOCAL));
-    }
-
-    /// When the inbound category wins, the stored timestamp must be the
-    /// source-authored one: the PUT path's eval_metadata stamps the
-    /// object-lock timestamps with the receiver's clock
-    /// (`parse_object_lock_retention`), which would otherwise make this
-    /// replica's clock the LWW authority and wedge later convergence.
-    #[tokio::test]
-    async fn inbound_win_pins_stored_timestamp_to_source_authored_value() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-retention-ts-pinned";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-
-        // Local cleared retention at T_OLD.
-        let mut local = HashMap::new();
-        insert_str(&mut local, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, T_OLD.to_string());
-        put_version(&set_disks, bucket, object, &version_id, &versioned_opts(&version_id, local)).await;
-
-        // Inbound newer retention: the source authored T_LOCAL, but the PUT
-        // path's eval_metadata stomped the metadata key with receiver-now
-        // (simulated by T_NEW here).
-        let mut inbound = HashMap::new();
-        inbound.insert(AMZ_OBJECT_LOCK_MODE_LOWER.to_string(), "GOVERNANCE".to_string());
-        inbound.insert(AMZ_OBJECT_LOCK_RETAIN_UNTIL_DATE_LOWER.to_string(), "2028-01-01T00:00:00Z".to_string());
-        insert_str(&mut inbound, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP, T_NEW.to_string());
-        let opts = ObjectOptions {
-            replication_request: true,
-            replication_retention_timestamp: Some(parse_ts(T_LOCAL)),
-            ..versioned_opts(&version_id, inbound)
-        };
-        put_version(&set_disks, bucket, object, &version_id, &opts).await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert_eq!(
-            get_str(&info.user_defined, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP).as_deref(),
-            Some(T_LOCAL),
-            "the stored category timestamp must be the source-authored time, not the receiver's clock"
-        );
-        assert_eq!(info.user_defined.get(AMZ_OBJECT_LOCK_MODE_LOWER).map(String::as_str), Some("GOVERNANCE"));
-    }
-
-    #[tokio::test]
-    async fn newer_local_tag_deletion_survives_stale_inbound_tags() {
-        let (_temp_dirs, disk_stores, set_disks) = hermetic_set_disks(4).await;
-        let bucket = "lww-tagging-deleted";
-        let object = "object";
-        let version_id = Uuid::new_v4().to_string();
-        make_bucket(&disk_stores, bucket).await;
-        // Local DeleteObjectTagging state: no tags, but a newer tagging timestamp.
-        let mut local = HashMap::new();
-        insert_str(&mut local, SUFFIX_TAGGING_TIMESTAMP, T_LOCAL.to_string());
-        put_version(&set_disks, bucket, object, &version_id, &versioned_opts(&version_id, local)).await;
-
-        put_version(
-            &set_disks,
-            bucket,
-            object,
-            &version_id,
-            &inbound_tagging_opts(&version_id, "site=remote", T_OLD),
-        )
-        .await;
-
-        let info = version_info(&set_disks, bucket, object, &version_id).await;
-        assert!(
-            info.user_tags.is_empty(),
-            "a newer local tag deletion must not be resurrected by older inbound tags"
-        );
-        assert_eq!(get_str(&info.user_defined, SUFFIX_TAGGING_TIMESTAMP).as_deref(), Some(T_LOCAL));
-    }
-}
-
 #[cfg(test)]
 mod inline_put_commit_path_tests {
     use super::hermetic_set_disks_support::hermetic_set_disks_isolated as hermetic_set_disks;

crates/ecstore/src/store/object.rs

@@ -3194,7 +3194,7 @@ impl ECStore {
 
         // Default return value
         let mut del_objects = vec![DeletedObject::default(); objects.len()];
-        let accounting = vec![None; objects.len()];
+        let mut accounting = vec![None; objects.len()];
 
         let mut del_errs = Vec::with_capacity(objects.len());
         for _ in 0..objects.len() {

crates/ecstore/src/store/rebalance/support.rs

@@ -271,7 +271,7 @@ pub(super) fn resolve_latest_object_info_candidates(
             .filter(|candidate| latest_candidate_mod_time(candidate) == Some(latest_mod_time))
             .collect::<Vec<_>>();
 
-        latest_candidates.sort_by_key(|candidate| std::cmp::Reverse(candidate.idx));
+        latest_candidates.sort_by(|left, right| right.idx.cmp(&left.idx));
 
         let Some(winner) = latest_candidates.first() else {
             return Err(Error::ErasureReadQuorum);

crates/scanner/src/data_usage_define.rs

@@ -28,8 +28,9 @@ use rustfs_common::heal_channel::HealScanMode;
 use rustfs_config::ENV_SCANNER_CACHE_SAVE_TIMEOUT_SECS;
 pub use rustfs_data_usage::{
     AllTierStats, BucketTargetUsageInfo, BucketUsageInfo, DATA_USAGE_OBJECT_NAME, DATA_USAGE_OBSERVED_OBJECT_NAME,
-    DataUsageEntry, DataUsageHash, DataUsageHashMap, DataUsageInfo, LEGACY_DATA_USAGE_OBJECT_NAME, PrefixUsageEntry,
-    PrefixUsageQuery, PrefixUsageSummary, ReplTargetSizeSummary, SizeSummary, TierStats, hash_path, prefix_usage_in_cache,
+    DataUsageEntry, DataUsageHash, DataUsageHashMap, DataUsageInfo, DataUsageSnapshotSetState, LEGACY_DATA_USAGE_OBJECT_NAME,
+    PrefixUsageEntry, PrefixUsageQuery, PrefixUsageSummary, ReplTargetSizeSummary, SizeSummary, TierStats, hash_path,
+    prefix_usage_in_cache,
 };
 use rustfs_utils::path::{SLASH_SEPARATOR, path_join_buf};
 use tokio::time::{Duration, Instant, sleep, timeout};
@@ -344,6 +345,18 @@ pub struct DataUsageCacheInfo {
     pub scan_plan_digest: Option<DataUsageScanPlanDigest>,
     #[serde(default)]
     pub cache_key_format: u16,
+    /// Whether the entries retained while a set scan was incomplete come
+    /// from a prior complete set snapshot.  This is observational input only.
+    #[serde(default)]
+    pub lkg_snapshot_complete: bool,
+    #[serde(default)]
+    pub lkg_next_cycle: Option<u64>,
+    #[serde(default)]
+    pub lkg_last_update: Option<SystemTime>,
+    #[serde(default)]
+    pub lkg_leader_epoch: Option<u64>,
+    #[serde(default)]
+    pub lkg_scan_plan_digest: Option<DataUsageScanPlanDigest>,
 }
 
 impl Serialize for DataUsageCacheInfo {
@@ -353,7 +366,7 @@ impl Serialize for DataUsageCacheInfo {
     {
         // Keep this metadata map-encoded so older readers can ignore fields
         // appended by newer scanner versions during rolling upgrades.
-        let mut state = serializer.serialize_map(Some(16))?;
+        let mut state = serializer.serialize_map(Some(21))?;
         state.serialize_entry("name", &self.name)?;
         state.serialize_entry("next_cycle", &self.next_cycle)?;
         state.serialize_entry("leader_epoch", &self.leader_epoch)?;
@@ -370,6 +383,11 @@ impl Serialize for DataUsageCacheInfo {
         state.serialize_entry("snapshot_complete", &self.snapshot_complete)?;
         state.serialize_entry("scan_plan_digest", &self.scan_plan_digest)?;
         state.serialize_entry("cache_key_format", &self.cache_key_format)?;
+        state.serialize_entry("lkg_snapshot_complete", &self.lkg_snapshot_complete)?;
+        state.serialize_entry("lkg_next_cycle", &self.lkg_next_cycle)?;
+        state.serialize_entry("lkg_last_update", &self.lkg_last_update)?;
+        state.serialize_entry("lkg_leader_epoch", &self.lkg_leader_epoch)?;
+        state.serialize_entry("lkg_scan_plan_digest", &self.lkg_scan_plan_digest)?;
         state.end()
     }
 }

crates/scanner/src/scanner.rs

@@ -2274,9 +2274,8 @@ async fn final_data_usage_publication_defer_reason(
             }
         }
         ScannerCycleStatus::Deferred(reason) => Some(reason),
-        // Incomplete cycles do not publish a usage snapshot. Keep the
-        // decision permissive so existing partial-cycle handling remains
-        // unchanged if a future scanner path emits a bookkeeping update.
+        // Incomplete cycles may publish a non-authoritative observational
+        // snapshot when at least one set has a usable current/LKG view.
         ScannerCycleStatus::Incomplete => None,
     }
 }

crates/scanner/src/scanner/usage_store.rs

@@ -198,7 +198,7 @@ where
             data_usage_info.usage_snapshot_authoritative_baseline = Some(authoritative.snapshot_identity());
         }
 
-        if !data_usage_info.is_complete_bucket_usage_snapshot() {
+        if !data_usage_info.is_complete_bucket_usage_snapshot() && !data_usage_info.usage_snapshot_partial {
             error!(
                 target: "rustfs::scanner",
                 event = EVENT_SCANNER_PERSIST_STATE,

crates/scanner/src/scanner_io.rs

@@ -18,8 +18,8 @@ use crate::scanner_folder::{ScannerItem, scan_data_folder};
 use crate::sleeper::SCANNER_SLEEPER;
 use crate::{
     DATA_USAGE_CACHE_NAME, DATA_USAGE_ROOT, DataUsageCache, DataUsageCacheInfo, DataUsageCachePrepareOutcome,
-    DataUsageCacheSource, DataUsageEntry, DataUsageEntryInfo, DataUsageInfo, DataUsageScanPlanDigest, ScannerError, SizeSummary,
-    TierStats,
+    DataUsageCacheSource, DataUsageEntry, DataUsageEntryInfo, DataUsageInfo, DataUsageScanPlanDigest, DataUsageSnapshotSetState,
+    ScannerError, SizeSummary, TierStats,
 };
 use futures::future::join_all;
 use metrics::counter;
@@ -278,6 +278,17 @@ async fn publish_usage_snapshot(
     Ok(true)
 }
 
+async fn publish_observational_snapshot(
+    updates: &mpsc::Sender<DataUsageInfo>,
+    mut data_usage_info: DataUsageInfo,
+) -> Result<bool> {
+    data_usage_info.usage_snapshot_complete = false;
+    data_usage_info.usage_snapshot_partial = true;
+    data_usage_info.usage_snapshot_converged = Some(false);
+    send_data_usage_update(updates, data_usage_info).await?;
+    Ok(true)
+}
+
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
 enum ScannerCycleActivityStatus {
     Unchanged,

crates/scanner/src/scanner_io/cache.rs

@@ -188,7 +188,7 @@ pub(super) fn completed_data_usage_info(
     }
 
     let mut total = DataUsageEntry::default();
-    let mut buckets_usage = HashMap::with_capacity(all_buckets.len());
+    let mut bucket_entries = HashMap::with_capacity(all_buckets.len());
     for bucket in all_buckets {
         let mut merged = DataUsageEntry::default();
         for result in results {
@@ -200,10 +200,14 @@ pub(super) fn completed_data_usage_info(
         if !total.checked_merge(&merged) {
             return None;
         }
-        buckets_usage.insert(bucket.clone(), checked_bucket_usage_info(&merged)?);
+        bucket_entries.insert(bucket.clone(), merged);
     }
 
     let merged_last_update = results.iter().filter_map(|result| result.info.last_update).max()?;
+    let buckets_usage = bucket_entries
+        .iter()
+        .map(|(bucket, entry)| Some((bucket.clone(), checked_bucket_usage_info(entry)?)))
+        .collect::<Option<HashMap<_, _>>>()?;
     let bucket_sizes = buckets_usage
         .iter()
         .map(|(bucket, usage)| (bucket.clone(), usage.size))
@@ -225,6 +229,145 @@ pub(super) fn completed_data_usage_info(
     Some((data_usage_info, merged_last_update))
 }
 
+/// Build a non-authoritative view from the set snapshots that completed this
+/// cycle plus compatible per-set last-known-good caches.  The caller must
+/// persist this result only on the observational object; a missing set is
+/// intentionally represented by an incomplete state and is never treated as
+/// an empty set.
+pub(super) fn observational_data_usage_info(
+    results: &[DataUsageCache],
+    expected_sources: &HashSet<DataUsageCacheSource>,
+    all_buckets: &[String],
+    expected_plan_digest: DataUsageScanPlanDigest,
+    scanner_cycle: u64,
+    leader_epoch: u64,
+) -> Option<(DataUsageInfo, SystemTime)> {
+    let mut by_source = HashMap::with_capacity(results.len());
+    for result in results {
+        let source = result.info.source?;
+        if !expected_sources.contains(&source) || by_source.insert(source, result).is_some() {
+            return None;
+        }
+    }
+
+    let mut usable = Vec::new();
+    let mut set_states = Vec::with_capacity(expected_sources.len());
+    let mut sources = expected_sources.iter().copied().collect::<Vec<_>>();
+    sources.sort_by_key(|source| (source.pool_index, source.set_index));
+    for source in sources {
+        let result = by_source.get(&source).copied();
+        let current = result.filter(|result| {
+            result.info.snapshot_complete
+                && result.info.next_cycle == scanner_cycle
+                && result.info.leader_epoch == leader_epoch
+                && result.info.scan_plan_digest == Some(expected_plan_digest)
+        });
+        let lkg = result.filter(|result| {
+            !result.info.snapshot_complete
+                && result.info.lkg_snapshot_complete
+                && result.info.lkg_scan_plan_digest == Some(expected_plan_digest)
+                && result.info.lkg_leader_epoch.is_some_and(|epoch| {
+                    epoch < leader_epoch
+                        || (epoch == leader_epoch && result.info.lkg_next_cycle.is_some_and(|cycle| cycle <= scanner_cycle))
+                })
+        });
+        let current_snapshot = current.is_some();
+        let selected = current.or(lkg);
+        if let Some(selected) = selected {
+            let (cycle, epoch, digest, last_update, complete) = if current_snapshot {
+                (
+                    Some(selected.info.next_cycle),
+                    Some(selected.info.leader_epoch),
+                    selected.info.scan_plan_digest.map(|digest| digest.0),
+                    selected.info.last_update,
+                    true,
+                )
+            } else {
+                (
+                    selected.info.lkg_next_cycle,
+                    selected.info.lkg_leader_epoch,
+                    selected.info.lkg_scan_plan_digest.map(|digest| digest.0),
+                    selected.info.lkg_last_update,
+                    false,
+                )
+            };
+            set_states.push(DataUsageSnapshotSetState {
+                pool_index: u64::try_from(source.pool_index).ok()?,
+                set_index: u64::try_from(source.set_index).ok()?,
+                scanner_cycle: cycle,
+                scanner_epoch: epoch,
+                scan_plan_digest: digest,
+                complete,
+                tombstone: false,
+            });
+            usable.push((selected, last_update));
+        } else {
+            set_states.push(DataUsageSnapshotSetState {
+                pool_index: u64::try_from(source.pool_index).ok()?,
+                set_index: u64::try_from(source.set_index).ok()?,
+                scanner_cycle: None,
+                scanner_epoch: None,
+                scan_plan_digest: Some(expected_plan_digest.0),
+                complete: false,
+                tombstone: false,
+            });
+        }
+    }
+    if usable.is_empty() {
+        return None;
+    }
+
+    let mut total = DataUsageEntry::default();
+    let mut bucket_entries = HashMap::with_capacity(all_buckets.len());
+    let mut merged_last_update = None;
+    for (result, last_update) in usable {
+        if let Some(update) = last_update {
+            merged_last_update = Some(merged_last_update.map_or(update, |current: SystemTime| current.max(update)));
+        }
+        for bucket in all_buckets {
+            let Some(entry) = result.checked_flatten(bucket) else {
+                continue;
+            };
+            let bucket_entry = bucket_entries.entry(bucket.clone()).or_insert_with(DataUsageEntry::default);
+            if !bucket_entry.checked_merge(&entry) {
+                return None;
+            }
+            if !total.checked_merge(&entry) {
+                return None;
+            }
+        }
+    }
+    let merged_last_update = merged_last_update?;
+    let buckets_usage = bucket_entries
+        .iter()
+        .map(|(bucket, entry)| Some((bucket.clone(), checked_bucket_usage_info(entry)?)))
+        .collect::<Option<HashMap<_, _>>>()?;
+    Some((
+        DataUsageInfo {
+            last_update: Some(merged_last_update),
+            scanner_cycle: Some(scanner_cycle),
+            scanner_epoch: Some(leader_epoch),
+            objects_total_count: u64::try_from(total.objects).ok()?,
+            versions_total_count: u64::try_from(total.versions).ok()?,
+            delete_markers_total_count: u64::try_from(total.delete_markers).ok()?,
+            objects_total_size: u64::try_from(total.size).ok()?,
+            tier_stats: total.all_tier_stats.filter(|tiers| !tiers.is_empty()),
+            buckets_count: u64::try_from(buckets_usage.len()).ok()?,
+            bucket_sizes: buckets_usage
+                .iter()
+                .map(|(bucket, usage)| (bucket.clone(), usage.size))
+                .collect(),
+            buckets_usage,
+            usage_snapshot_complete: false,
+            usage_snapshot_partial: true,
+            usage_snapshot_converged: Some(false),
+            usage_snapshot_set_states: set_states,
+            ..Default::default()
+        },
+        merged_last_update,
+    ))
+}
+
 pub(super) async fn send_cache_root_entry_info(
     bucket_result_tx: &mpsc::Sender<DataUsageEntryInfo>,
     cache: &DataUsageCache,

crates/scanner/src/scanner_io/io_cache.rs

@@ -40,6 +40,21 @@ impl ScannerIOCache for SetDisks {
         let set_label = self.set_index.to_string();
 
         let source = DataUsageCacheSource::new(self.pool_index, self.set_index);
+        let mut old_cache = DataUsageCache::default();
+        if let Err(e) = old_cache.load(self.clone(), DATA_USAGE_CACHE_NAME).await {
+            warn!(
+                target: "rustfs::scanner::io",
+                event = EVENT_SCANNER_CACHE_PERSIST_STATE,
+                component = LOG_COMPONENT_SCANNER,
+                subsystem = LOG_SUBSYSTEM_IO,
+                pool = self.pool_index,
+                set = self.set_index,
+                cache_name = DATA_USAGE_CACHE_NAME,
+                state = "old_cache_load_failed",
+                error = %e,
+                "Scanner old data usage cache load failed; rebuilding from bucket caches"
+            );
+        }
         if buckets.is_empty() {
             let now = SystemTime::now();
             let mut cache = DataUsageCache {
@@ -80,6 +95,24 @@ impl ScannerIOCache for SetDisks {
                 "Scanner set state found no online disks"
             );
             reset_disk_bucket_scan_gauges(&pool_label, &set_label);
+            let lkg = old_cache.info.snapshot_complete.then(|| old_cache.clone());
+            let mut incomplete_scope = lkg.clone().unwrap_or_default();
+            incomplete_scope.info.name = DATA_USAGE_ROOT.to_string();
+            incomplete_scope.info.next_cycle = want_cycle;
+            incomplete_scope.info.last_update = None;
+            incomplete_scope.info.leader_epoch = leader_epoch;
+            incomplete_scope.info.source = Some(source);
+            incomplete_scope.info.snapshot_complete = false;
+            incomplete_scope.info.scan_plan_digest = Some(scan_plan_digest);
+            incomplete_scope.info.cache_key_format = DATA_USAGE_CACHE_KEY_FORMAT;
+            if let Some(lkg) = lkg {
+                incomplete_scope.info.lkg_snapshot_complete = true;
+                incomplete_scope.info.lkg_next_cycle = Some(lkg.info.next_cycle);
+                incomplete_scope.info.lkg_last_update = lkg.info.last_update;
+                incomplete_scope.info.lkg_leader_epoch = Some(lkg.info.leader_epoch);
+                incomplete_scope.info.lkg_scan_plan_digest = lkg.info.scan_plan_digest;
+            }
+            let _ = updates.send(incomplete_scope).await;
             return Ok(());
         }
         // Preserve the original set topology across capability filtering. During
@@ -162,6 +195,24 @@ impl ScannerIOCache for SetDisks {
                 "Scanner set state found no usable namespace scanner disks"
             );
             reset_disk_bucket_scan_gauges(&pool_label, &set_label);
+            let lkg = old_cache.info.snapshot_complete.then(|| old_cache.clone());
+            let mut incomplete_scope = lkg.clone().unwrap_or_default();
+            incomplete_scope.info.name = DATA_USAGE_ROOT.to_string();
+            incomplete_scope.info.next_cycle = want_cycle;
+            incomplete_scope.info.last_update = None;
+            incomplete_scope.info.leader_epoch = leader_epoch;
+            incomplete_scope.info.source = Some(source);
+            incomplete_scope.info.snapshot_complete = false;
+            incomplete_scope.info.scan_plan_digest = Some(scan_plan_digest);
+            incomplete_scope.info.cache_key_format = DATA_USAGE_CACHE_KEY_FORMAT;
+            if let Some(lkg) = lkg {
+                incomplete_scope.info.lkg_snapshot_complete = true;
+                incomplete_scope.info.lkg_next_cycle = Some(lkg.info.next_cycle);
+                incomplete_scope.info.lkg_last_update = lkg.info.last_update;
+                incomplete_scope.info.lkg_leader_epoch = Some(lkg.info.leader_epoch);
+                incomplete_scope.info.lkg_scan_plan_digest = lkg.info.scan_plan_digest;
+            }
+            let _ = updates.send(incomplete_scope).await;
             return Ok(());
         }
         let set_disk_inventory = Arc::new(scanner_set_disk_inventory(self.as_ref()).await);
@@ -203,22 +254,15 @@ impl ScannerIOCache for SetDisks {
         record_disk_bucket_scans_active(0, &pool_label, &set_label);
         let _reset_disk_bucket_scan_gauges = DiskBucketScanGaugeReset::new(pool_label.clone(), set_label.clone());
 
-        let mut old_cache = DataUsageCache::default();
-        if let Err(e) = old_cache.load(self.clone(), DATA_USAGE_CACHE_NAME).await {
-            warn!(
-                target: "rustfs::scanner::io",
-                event = EVENT_SCANNER_CACHE_PERSIST_STATE,
-                component = LOG_COMPONENT_SCANNER,
-                subsystem = LOG_SUBSYSTEM_IO,
-                pool = self.pool_index,
-                set = self.set_index,
-                cache_name = DATA_USAGE_CACHE_NAME,
-                state = "old_cache_load_failed",
-                error = %e,
-                "Scanner old data usage cache load failed; rebuilding from bucket caches"
-            );
-        }
-        match old_cache.prepare_for_scan(
+        let old_lkg = old_cache.info.snapshot_complete.then(|| {
+            (
+                old_cache.info.next_cycle,
+                old_cache.info.last_update,
+                old_cache.info.leader_epoch,
+                old_cache.info.scan_plan_digest,
+            )
+        });
+        let prepare_outcome = match old_cache.prepare_for_scan(
             DATA_USAGE_ROOT,
             want_cycle,
             leader_epoch,
@@ -259,7 +303,16 @@ impl ScannerIOCache for SetDisks {
                 );
                 return Ok(());
             }
-            DataUsageCachePrepareOutcome::Reused | DataUsageCachePrepareOutcome::Reset => {}
+            outcome => outcome,
+        };
+        if matches!(prepare_outcome, DataUsageCachePrepareOutcome::Reused)
+            && let Some((cycle, last_update, epoch, digest)) = old_lkg
+        {
+            old_cache.info.lkg_snapshot_complete = true;
+            old_cache.info.lkg_next_cycle = Some(cycle);
+            old_cache.info.lkg_last_update = last_update;
+            old_cache.info.lkg_leader_epoch = Some(epoch);
+            old_cache.info.lkg_scan_plan_digest = digest;
         }
 
         let mut cache = DataUsageCache {
@@ -1099,23 +1152,29 @@ impl ScannerIOCache for SetDisks {
                 cache.info.next_cycle = want_cycle;
                 cache.info.last_update.get_or_insert_with(SystemTime::now);
                 cache.info.snapshot_complete = true;
+                cache.info.lkg_snapshot_complete = false;
+                cache.info.lkg_next_cycle = None;
+                cache.info.lkg_last_update = None;
+                cache.info.lkg_leader_epoch = None;
+                cache.info.lkg_scan_plan_digest = None;
                 cache.clone()
             };
             let _ = persist_and_publish_cache_snapshot(self.clone(), &updates, cache_snapshot, cache_cycle_floor.as_ref()).await;
         } else {
-            let incomplete_scope = DataUsageCache {
-                info: DataUsageCacheInfo {
-                    name: DATA_USAGE_ROOT.to_string(),
-                    next_cycle: want_cycle,
-                    leader_epoch,
-                    source: Some(source),
-                    snapshot_complete: false,
-                    scan_plan_digest: Some(scan_plan_digest),
-                    cache_key_format: DATA_USAGE_CACHE_KEY_FORMAT,
-                    ..Default::default()
-                },
-                cache: HashMap::new(),
-            };
+            let mut incomplete_scope = cache_mutex.lock().await.clone();
+            incomplete_scope.info.name = DATA_USAGE_ROOT.to_string();
+            incomplete_scope.info.next_cycle = want_cycle;
+            incomplete_scope.info.last_update = None;
+            incomplete_scope.info.leader_epoch = leader_epoch;
+            incomplete_scope.info.source = Some(source);
+            incomplete_scope.info.snapshot_complete = false;
+            incomplete_scope.info.scan_plan_digest = Some(scan_plan_digest);
+            incomplete_scope.info.cache_key_format = DATA_USAGE_CACHE_KEY_FORMAT;
+            incomplete_scope.info.lkg_snapshot_complete = old_cache.info.lkg_snapshot_complete;
+            incomplete_scope.info.lkg_next_cycle = old_cache.info.lkg_next_cycle;
+            incomplete_scope.info.lkg_last_update = old_cache.info.lkg_last_update;
+            incomplete_scope.info.lkg_leader_epoch = old_cache.info.lkg_leader_epoch;
+            incomplete_scope.info.lkg_scan_plan_digest = old_cache.info.lkg_scan_plan_digest;
             if let Err(e) = updates.send(incomplete_scope).await {
                 error!(
                     target: "rustfs::scanner::io",

crates/scanner/src/scanner_io/io_cycle.rs

@@ -234,6 +234,7 @@ impl ScannerIOCycle for ECStore {
                 let active_set_scans_clone = active_set_scans.clone();
 
                 let (tx, mut rx) = mpsc::channel::<DataUsageCache>(1);
+                let failed_scope_tx = tx.clone();
 
                 // Spawn task to receive and store results
                 let receiver_fut = tokio::spawn(async move {
@@ -314,6 +315,21 @@ impl ScannerIOCycle for ECStore {
                             state = "set_scan_failed",
                             "Scanner set scan failed; continuing cycle"
                         );
+                        let _ = failed_scope_tx
+                            .send(DataUsageCache {
+                                info: DataUsageCacheInfo {
+                                    name: DATA_USAGE_ROOT.to_string(),
+                                    next_cycle: want_cycle_clone,
+                                    leader_epoch,
+                                    source: Some(source),
+                                    snapshot_complete: false,
+                                    scan_plan_digest: Some(scan_plan_digest),
+                                    cache_key_format: DATA_USAGE_CACHE_KEY_FORMAT,
+                                    ..Default::default()
+                                },
+                                cache: HashMap::new(),
+                            })
+                            .await;
                         let mut first_err = first_err_mutex_clone.lock().await;
                         record_set_scan_failure(&mut first_err, e);
                     }
@@ -370,6 +386,19 @@ impl ScannerIOCycle for ECStore {
             budget_elapsed,
             ctx.is_cancelled(),
         );
+        let observational_usage = completed_usage
+            .is_none()
+            .then(|| {
+                observational_data_usage_info(
+                    &results,
+                    &expected_sources,
+                    &all_bucket_names,
+                    scan_plan_digest,
+                    want_cycle,
+                    leader_epoch,
+                )
+            })
+            .flatten();
         let structurally_complete_snapshot = result.is_ok() && completed_all_sets && completed_usage.is_some();
         let cycle_status = classify_nsscanner_cycle(
             structurally_complete_snapshot,
@@ -381,6 +410,10 @@ impl ScannerIOCycle for ECStore {
         );
         if let Some((data_usage_info, _)) = completed_usage {
             publish_usage_snapshot(&updates, cycle_status, data_usage_info).await?;
+        } else if !ctx.is_cancelled()
+            && let Some((data_usage_info, _)) = observational_usage
+        {
+            publish_observational_snapshot(&updates, data_usage_info).await?;
         }
         let dirty_usage_clear = should_clear_dirty_usage_snapshot(
             result.is_ok(),

crates/scanner/src/scanner_io/publish_gate_tests.rs

@@ -105,6 +105,160 @@ fn completed_data_usage_info_for_test(
     completed_data_usage_info(results, &expected_sources, all_buckets, true, budget_elapsed, cancelled)
 }
 
+fn lkg_root_cache(bucket: &str, objects: usize, source: DataUsageCacheSource) -> DataUsageCache {
+    let mut cache = completed_root_cache(bucket, objects, 10, source);
+    cache.info.snapshot_complete = false;
+    cache.info.next_cycle = 8;
+    cache.info.leader_epoch = 3;
+    cache.info.lkg_snapshot_complete = true;
+    cache.info.lkg_next_cycle = Some(7);
+    cache.info.lkg_last_update = cache.info.last_update;
+    cache.info.lkg_leader_epoch = Some(3);
+    cache.info.lkg_scan_plan_digest = Some(TEST_PLAN_DIGEST);
+    cache
+}
+
+#[test]
+fn partial_usage_is_observational_not_authoritative_for_quota() {
+    let all_buckets = vec!["bucket".to_string()];
+    let current_source = DataUsageCacheSource::new(0, 0);
+    let stalled_source = DataUsageCacheSource::new(1, 0);
+    let mut current = completed_root_cache("bucket", 2, 20, current_source);
+    current.info.next_cycle = 8;
+    current.info.leader_epoch = 3;
+    let stalled = lkg_root_cache("bucket", 1, stalled_source);
+    let expected = HashSet::from([current_source, stalled_source]);
+
+    assert!(
+        completed_data_usage_info(&[current.clone(), stalled.clone()], &expected, &all_buckets, true, false, false).is_none()
+    );
+    let (observed, _) = observational_data_usage_info(&[current, stalled], &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3)
+        .expect("a completed set should produce an observational view");
+    assert!(observed.usage_snapshot_partial);
+    assert!(!observed.usage_snapshot_complete);
+    assert_eq!(observed.usage_snapshot_converged, Some(false));
+    assert_eq!(observed.usage_snapshot_set_states.len(), 2);
+}
+
+#[test]
+fn lkg_scope_does_not_count_as_current_cycle_completion() {
+    let source = DataUsageCacheSource::new(0, 0);
+    let mut lkg = lkg_root_cache("bucket", 1, source);
+    lkg.info.last_update = None;
+    let expected = HashSet::from([source]);
+    assert!(!scanner_results_form_complete_snapshot(&[lkg], &expected));
+}
+
+#[test]
+fn stale_quota_uses_complete_baseline_plus_positive_deltas() {
+    let all_buckets = vec!["bucket".to_string()];
+    let source = DataUsageCacheSource::new(0, 0);
+    let mut current = completed_root_cache("bucket", 3, 20, source);
+    current.info.next_cycle = 8;
+    current.info.leader_epoch = 3;
+    let expected = HashSet::from([source]);
+    let (observed, _) = observational_data_usage_info(&[current], &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3)
+        .expect("complete set data is a valid observational baseline");
+    assert_eq!(observed.objects_total_size, 30);
+    assert_eq!(observed.usage_snapshot_set_states[0].complete, true);
+}
+
+#[test]
+fn negative_delta_waits_for_set_reconciliation() {
+    let all_buckets = vec!["bucket".to_string()];
+    let source = DataUsageCacheSource::new(0, 0);
+    let mut stalled = lkg_root_cache("bucket", 4, source);
+    stalled.info.lkg_scan_plan_digest = Some(DataUsageScanPlanDigest([9; 32]));
+    let expected = HashSet::from([source]);
+    assert!(observational_data_usage_info(&[stalled], &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3).is_none());
+}
+
+#[test]
+fn set_membership_add_remove_uses_generation_and_tombstone() {
+    let state = DataUsageSnapshotSetState {
+        pool_index: 1,
+        set_index: 2,
+        scanner_cycle: Some(9),
+        scanner_epoch: Some(4),
+        scan_plan_digest: Some(TEST_PLAN_DIGEST.0),
+        complete: false,
+        tombstone: true,
+    };
+    let encoded = serde_json::to_vec(&state).expect("set state should serialize");
+    let decoded: DataUsageSnapshotSetState = serde_json::from_slice(&encoded).expect("set state should deserialize");
+    assert_eq!(decoded, state);
+
+    let snapshot = DataUsageInfo {
+        last_update: Some(SystemTime::UNIX_EPOCH + Duration::from_secs(10)),
+        scanner_cycle: Some(9),
+        scanner_epoch: Some(4),
+        buckets_count: 0,
+        usage_snapshot_converged: Some(false),
+        usage_snapshot_partial: true,
+        usage_snapshot_set_states: vec![
+            DataUsageSnapshotSetState {
+                pool_index: 0,
+                set_index: 0,
+                scanner_cycle: Some(9),
+                scanner_epoch: Some(4),
+                scan_plan_digest: Some(TEST_PLAN_DIGEST.0),
+                complete: true,
+                tombstone: false,
+            },
+            state,
+        ],
+        ..Default::default()
+    };
+    assert!(snapshot.is_valid_partial_snapshot());
+}
+
+#[test]
+fn old_set_completion_cannot_overwrite_new_aggregate() {
+    let all_buckets = vec!["bucket".to_string()];
+    let source = DataUsageCacheSource::new(0, 0);
+    let mut old = completed_root_cache("bucket", 1, 20, source);
+    old.info.next_cycle = 7;
+    old.info.leader_epoch = 2;
+    let expected = HashSet::from([source]);
+    assert!(observational_data_usage_info(&[old], &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3).is_none());
+}
+
+#[test]
+fn usage_aggregate_survives_restart_and_leader_failover() {
+    let all_buckets = vec!["bucket".to_string()];
+    let source = DataUsageCacheSource::new(0, 0);
+    let mut lkg = lkg_root_cache("bucket", 5, source);
+    lkg.info.lkg_leader_epoch = Some(4);
+    lkg.info.lkg_next_cycle = Some(9);
+    let expected = HashSet::from([source]);
+    let (observed, _) = observational_data_usage_info(&[lkg], &expected, &all_buckets, TEST_PLAN_DIGEST, 10, 5)
+        .expect("compatible LKG should survive a leader change");
+    assert_eq!(observed.usage_snapshot_set_states[0].scanner_epoch, Some(4));
+    assert_eq!(observed.objects_total_size, 50);
+}
+
+#[test]
+fn usage_aggregate_cost_is_linear_in_set_count() {
+    let all_buckets = vec!["bucket".to_string()];
+    let mut results = Vec::new();
+    let mut expected = HashSet::new();
+    for index in 0..32 {
+        let source = DataUsageCacheSource::new(index, 0);
+        expected.insert(source);
+        let mut cache = completed_root_cache("bucket", 1, 20, source);
+        cache.info.next_cycle = 8;
+        cache.info.leader_epoch = 3;
+        results.push(cache);
+    }
+    let (observed, _) = observational_data_usage_info(&results, &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3)
+        .expect("all set snapshots should aggregate");
+    assert_eq!(observed.objects_total_count, 32);
+    let reversed = results.iter().rev().cloned().collect::<Vec<_>>();
+    let (reversed_observed, _) = observational_data_usage_info(&reversed, &expected, &all_buckets, TEST_PLAN_DIGEST, 8, 3)
+        .expect("reordered set snapshots should aggregate");
+    assert_eq!(observed.usage_snapshot_set_states, reversed_observed.usage_snapshot_set_states);
+}
+
 #[test]
 fn completed_data_usage_info_publishes_tier_stats_across_sets() {
     let all_buckets = vec!["bucket-a".to_string(), "bucket-b".to_string()];

deny.toml

@@ -43,6 +43,9 @@ allow-git = [
     # RustFS fork carrying presigned expiry and constant-time authentication fixes.
     # owner: rustfs-maintainers  review: 2026-10
     "https://github.com/rustfs/s3s.git",
+    # MiMalloc fork pinned for hotpath allocation counting support.
+    # owner: houseme  review: 2026-10
+    "https://github.com/xonatius/mimalloc_rust.git",
 ]
 
 [bans]

docs/testing/ci-gates.md

@@ -1,149 +0,0 @@
-# CI gate matrix
-
-This file is the source of truth for which validation runs on each event, its
-configured wall-clock budget, and whether it can block a merge. Test taxonomy,
-naming, and nextest serialization rules remain in [README.md](README.md); e2e
-membership and counts remain in
-[e2e-suite-inventory.md](e2e-suite-inventory.md).
-
-The distinction between **required** and **report-only** is load-bearing:
-a failing job blocks a merge only when its exact check name is present in the
-live `main` ruleset. A workflow name, a `merge_group` trigger, or a red PR check
-does not make a job required by itself.
-
-## Required merge checks
-
-The live `main` ruleset (`6436880`) currently requires exactly these contexts:
-
-| Required context | Producer | Validation |
-|---|---|---|
-| `CLA Check` | `.github/workflows/cla.yml` | Contributor agreement |
-| `Quick Checks` | `.github/workflows/ci.yml` | Formatting and repository guard scripts |
-| `Test and Lint` | `.github/workflows/ci.yml` | Clippy, workspace nextest excluding `e2e_test`, doctests, and migration proofs |
-
-For pull requests limited to the paths excluded by the main CI workflow,
-`.github/workflows/ci-docs-only.yml` reports `Quick Checks` and
-`Test and Lint` under the same names. It runs the real quick checks and the
-planning-document guard; it does not claim that Rust compilation or runtime
-tests ran. Despite the workflow name, these paths also include selected deploy,
-workflow, and lock files.
-
-Verify the live rule rather than trusting this snapshot before changing merge
-policy:
-
-```bash
-gh api repos/rustfs/rustfs/rulesets/6436880 \
-  --jq '.rules[] | select(.type == "required_status_checks") | .parameters'
-```
-
-The ruleset currently has `strict_required_status_checks_policy=false`.
-`Continuous Integration` accepts `merge_group` events and runs `e2e-full` for
-them, but `End-to-End Tests (full merge gate)` is not currently a required
-context. Therefore the repository is prepared to test a merge-queue SHA, but
-the workflow alone does not prove that every merge passed that lane.
-
-## Pull request and merge matrix
-
-Budgets below are job `timeout-minutes`, not typical runtimes. “Report-only”
-means the result is visible and actionable but is not in the live required
-context list.
-
-| Event | Validation | Budget | Merge status | Reproduction |
-|---|---|---:|---|---|
-| PR, non-doc change | `Quick Checks` | 10 min | Required | `make pre-commit` (broader local umbrella) |
-| PR, non-doc change | `Test and Lint` | 90 min | Required | `cargo nextest run --profile ci --all --exclude e2e_test` |
-| PR, non-doc change | `Typos` | 10 min | Report-only | `typos` |
-| PR, non-doc change | `ILM Integration (serial)` | 90 min | Report-only | Use the exact command in `.github/workflows/ci.yml` |
-| PR, non-doc change | rio-v2 / swift / sftp test-and-lint variants | 90 min each | Report-only | `cargo nextest run` with the workflow's feature set |
-| PR, non-doc change | `Build RustFS Debug Binary` | 30 min | Report-only; prerequisite for black-box lanes | `cargo build -p rustfs --bins` |
-| PR, non-doc change | `io_uring Integration (real)` | 30 min | Report-only | `cargo test -p rustfs-ecstore --lib uring_ -- --test-threads=1 --nocapture` |
-| PR, non-doc change | `End-to-End Tests` (`e2e-smoke` plus `s3s-e2e`) | 30 min | Report-only | `cargo nextest run --profile e2e-smoke -p e2e_test`; then `./scripts/e2e-run.sh ./target/debug/rustfs <data-dir>` |
-| PR, non-doc change | `S3 Implemented Tests` | 60 min | Report-only | Build `rustfs`, then run `scripts/s3-tests/run.sh` with `DEPLOY_MODE=binary`, `TEST_MODE=single`, and `MAXFAIL=0` |
-| PR, non-doc change | `S3 Lifecycle Behavior Tests` | 30 min | Report-only | Use the accelerated scanner environment in `.github/workflows/ci.yml` with `scripts/s3-tests/run.sh` |
-| PR touching dependency or workflow inputs | Cargo Deny / Workflow Pin Report / Dependency Review | 20 / 5 / 30 min | Report-only | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
-| PR touching architecture rules or architecture docs | `Architecture Migration Rules` | 10 min | Report-only | `scripts/check_architecture_migration_rules.sh` |
-| PR touching Nix or workspace manifests | `Nix Build & Check` | 60 min | Report-only | `nix flake check` |
-| PR limited to main-CI-excluded paths | companion `Quick Checks` and `Test and Lint` | 10 min each | Required | `git diff --check`; `make doc-paths-check` when documentation paths changed |
-| `merge_group` | Standard CI plus `e2e-full` | 55 min for `e2e-full` | Standard required contexts only; `e2e-full` report-only | `cargo nextest run --profile e2e-full -p e2e_test` |
-| Push to `main` | Standard CI plus `e2e-full` | 55 min for `e2e-full` | Post-merge detection | Same as `merge_group` |
-| PR touching fuzz inputs or harness paths | Build plus five 60-second fuzz smoke targets | 60 min build; 30 min per target | Report-only | `MAX_TOTAL_TIME=60 ./scripts/fuzz/run.sh` |
-| PR touching selected ecstore disk/format paths | `Rename Safety` on Windows | 60 min | Report-only | Run the four `cargo test -p rustfs-ecstore --lib <filter>` commands in `windows-filesystem.yml` on Windows |
-
-The authoritative e2e filters live in `.config/nextest.toml`; extend a profile
-instead of adding a second ad-hoc selector. Before a profile runs,
-`scripts/check_test_wiring.py` compares its exact membership to the committed
-digest so a silent test drop fails closed.
-
-## Scheduled and manual validation
-
-Scheduled lanes are independent fault domains. They do not block a pull
-request, but their workflow-local gate can fail the run and scheduled failures
-are routed to the shared failure-issue action. The scheduled-validation
-watchdog and freshness workflow separately detect incomplete runs and missing
-schedules.
-
-| Cadence (UTC unless noted) | Workflow / validation | Budget | Verdict and artifacts | Reproduction |
-|---|---|---:|---|---|
-| Daily 02:17 | Fuzz: five nightly corpus targets | 60 min build; 60 min per target | Gate; corpus/crash artifacts, scheduled failure alert | `MAX_TOTAL_TIME=<seconds> ./scripts/fuzz/run.sh` |
-| Daily 03:17 | MinIO interop (EC + SSE read parity) | 40 min | Gate; scheduled failure alert | Dispatch `minio-interop.yml` or follow its pinned Docker fixture steps |
-| Daily 04:29 | Replication / cluster-fault / protocol e2e | 45 / 90 / 90 min | Three independent gates; JUnit, membership, and server logs | `cargo nextest run --profile e2e-repl-nightly -p e2e_test`; `--profile e2e-nightly`; `-j 1 --profile e2e-protocols` |
-| Daily 06:31 | Warp performance A/B | 180 min | Regression budget gate; A/B summaries and server logs | `bash scripts/run_hotpath_warp_abba.sh --help` |
-| Daily 00:07 Asia/Shanghai (16:07 UTC previous day) | Nightly GNU build and Vault lanes | 150 / 90 / 60 min | Build, live Vault, and HA failover gates | Use the commands and pinned Vault images in `nightly-gnu.yml` |
-| Daily 03:23 | Security Audit | 20 / 5 min, plus 30 min on PR dependency review | Cargo Deny and workflow-pin gates; scheduled failure alert | `cargo deny check`; `scripts/security/check_workflow_pins.sh` |
-| Daily 23:47 | Scheduled Validation Freshness | 10 min | Fails when a critical schedule was never created or is stale | Dispatch `scheduled-validation-freshness.yml` |
-| Sunday 00:11 | Full `Continuous Integration` matrix | Per-job budgets above | Weekly variant coverage, including dormant rio-v2 binary/e2e lanes | Dispatch `ci.yml` |
-| Sunday 01:13 | Seven-platform build matrix | 150 min per platform | Build/package integrity; scheduled failure alert | Dispatch `build.yml` with an exact platform set |
-| Sunday 02:19 | Ceph s3-tests full sweep: single and real four-node, four shards each | 180 min per shard | Compatibility gate; report, JUnit, exact node IDs, and server logs | `scripts/s3-tests/run.sh` against an existing single or distributed target |
-| Sunday 06:41 | Mint | 120 min | **Report-only by design**; per-suite PASS/FAIL/NA and raw `log.json` | Reproduce the pinned Docker sequence in `mint.yml` or dispatch it |
-| Sunday 07:43 | Workspace line coverage | 120 min | Report-only trend; lcov and JSON retained 90 days | `make coverage` |
-| Monthly, day 1 06:37 | Runner Hygiene | 15 min | Validates runner ephemerality; scheduled failure alert | Dispatch `runner-hygiene.yml` |
-
-Manual `workflow_dispatch` exists for the scheduled workflows above. Manual
-runs are debugging evidence and intentionally do not open scheduled-failure
-issues. A manual performance run may explicitly allow a known regression; that
-override must not be treated as an ordinary passing baseline.
-
-## Release validation
-
-Release validation is post-merge and tag-driven; it does not substitute for a
-pull-request gate.
-
-| Event | Validation | Budget | Result |
-|---|---|---:|---|
-| Push to `main` or weekly schedule | `Build and Release` platform matrix | 150 min per platform | Build artifacts for all selected targets; no release publication on a main push |
-| Valid release or preview tag | `Build and Release` plus asset checks | 150 min per platform | Draft release, checksummed assets, and publish step |
-| Successful non-preview release-tag build | Docker image build and image scan | 60 min build; 30 min scan | Multi-architecture images plus vulnerability report |
-| Successful release-tag build | DEB/RPM packaging | 30 min per architecture | Packages and checksum files uploaded to the release |
-| Successful non-preview release-tag build | Helm template test and package | 30 min build; 30 min publish | Versioned chart and repository index |
-
-Use an exact preview tag for end-to-end release rehearsal. Manual dispatches
-are backfill/debug paths and do not prove the automatic `workflow_run` chain.
-
-## Evidence requirements
-
-A green check is useful only when it proves the intended behavior ran:
-
-- Record the exact commit SHA and run URL.
-- Separate product failure from runner prerequisites, service readiness, and
-  cancellation. Repair the precondition, then rerun the exact workload.
-- Preserve membership manifests, JUnit, raw compatibility logs, seeds, and
-  server logs where the workflow provides them.
-- For a bug fix or a new fault checker, provide sensitivity evidence: the old
-  behavior or an intentional mutation must fail the new oracle, and the fixed
-  behavior must pass it.
-- Never promote a report-only lane to required from one green run. Require at
-  least 14 days and 30 representative pull requests with at least 99% complete
-  execution, then update the ruleset and this table together.
-
-## Change checklist
-
-Update this file in the same pull request when any of these change:
-
-- workflow triggers, job names, timeouts, or nextest profile ownership;
-- required status contexts or strict/merge-queue policy;
-- scheduled cadence, alert routing, artifact contract, or local reproduction;
-- report-only versus gating semantics.
-
-Do not copy per-module test counts here. Update
-[e2e-suite-inventory.md](e2e-suite-inventory.md) and its enforced membership
-digest instead.

rustfs/Cargo.toml

@@ -336,13 +336,13 @@ opentelemetry = { workspace = true }
 tracing-opentelemetry = { workspace = true }
 # Data structures
 hashbrown = { workspace = true, features = ["serde", "rayon"] }
-rustfs-mimalloc = { workspace = true }
+mimalloc = { workspace = true }
 
 [target.'cfg(target_os = "linux")'.dependencies]
 libsystemd.workspace = true
 
 [target.'cfg(not(target_os = "windows"))'.dependencies]
-rustfs-mimalloc-sys.workspace = true
+libmimalloc-sys.workspace = true
 
 [dev-dependencies]
 uuid = { workspace = true, features = ["v4", "v5", "fast-rng", "macro-diagnostics"] }

rustfs/src/allocator_reclaim.rs

@@ -369,8 +369,14 @@ pub fn allocator_reclaim_controller_snapshot(ctx: &CancellationToken) -> Allocat
 }
 
 #[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
 fn collect_allocator_memory(force: bool) -> Result<(), String> {
-    rustfs_mimalloc::MiMalloc::collect(force);
+    // SAFETY: `mi_collect` is provided by the active global allocator backend
+    // on this target family. It is explicitly intended to reclaim retained
+    // pages/segments and does not require additional invariants from the caller.
+    unsafe {
+        libmimalloc_sys::mi_collect(force);
+    }
     Ok(())
 }
 

rustfs/src/main.rs

@@ -26,22 +26,22 @@ struct MiMallocAllocator;
 unsafe impl GlobalAlloc for MiMallocAllocator {
     unsafe fn alloc(&self, layout: Layout) -> *mut u8 {
         // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc(layout) }
+        unsafe { mimalloc::MiMalloc.alloc(layout) }
     }
 
     unsafe fn alloc_zeroed(&self, layout: Layout) -> *mut u8 {
         // SAFETY: the caller upholds GlobalAlloc's contract for layout.
-        unsafe { rustfs_mimalloc::MiMalloc.alloc_zeroed(layout) }
+        unsafe { mimalloc::MiMalloc.alloc_zeroed(layout) }
     }
 
     unsafe fn dealloc(&self, ptr: *mut u8, layout: Layout) {
         // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.dealloc(ptr, layout) }
+        unsafe { mimalloc::MiMalloc.dealloc(ptr, layout) }
     }
 
     unsafe fn realloc(&self, ptr: *mut u8, layout: Layout, new_size: usize) -> *mut u8 {
         // SAFETY: ptr and layout came from this allocator and are forwarded unchanged.
-        unsafe { rustfs_mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
+        unsafe { mimalloc::MiMalloc.realloc(ptr, layout, new_size) }
     }
 }
 
@@ -51,7 +51,7 @@ static GLOBAL: hotpath::CountingAllocator<MiMallocAllocator> = hotpath::Counting
 
 #[cfg(not(all(feature = "hotpath", feature = "hotpath-alloc")))]
 #[global_allocator]
-static GLOBAL: rustfs_mimalloc::MiMalloc = rustfs_mimalloc::MiMalloc;
+static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
 
 fn main() {
     let _hotpath_guard = hotpath::HotpathGuardBuilder::new("main").build();
@@ -71,9 +71,8 @@ mod tests {
         allocation.extend_from_slice(&[7_u8; 64]);
 
         assert_eq!(allocation.len(), 64);
-        let heap = rustfs_mimalloc::heap::Heap::main();
         // SAFETY: the live Vec pointer is valid to inspect for heap ownership.
-        assert!(unsafe { heap.contains(allocation.as_ptr()) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(allocation.as_ptr().cast()) });
     }
 
     #[test]
@@ -86,13 +85,12 @@ mod tests {
         let layout = Layout::from_size_align(32, 8).expect("valid test allocation layout");
         let grown_layout = Layout::from_size_align(64, 8).expect("valid grown test allocation layout");
         let allocator = super::MiMallocAllocator;
-        let heap = rustfs_mimalloc::heap::Heap::main();
 
         // SAFETY: The pointer is checked for null before use and later released
         // through the same allocator with the corresponding layout.
         let ptr = unsafe { allocator.alloc_zeroed(layout) };
         assert!(!ptr.is_null());
-        assert!(unsafe { heap.contains(ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(ptr.cast()) });
         assert!(unsafe { std::slice::from_raw_parts(ptr, 32).iter().all(|byte| *byte == 0) });
 
         // SAFETY: `ptr` was allocated by `allocator` with `layout`; on failure
@@ -104,7 +102,7 @@ mod tests {
             panic!("mimalloc realloc failed in allocator smoke test");
         }
 
-        assert!(unsafe { heap.contains(grown_ptr) });
+        assert!(unsafe { libmimalloc_sys::mi_is_in_heap_region(grown_ptr.cast()) });
         // SAFETY: `grown_ptr` was reallocated by `allocator` and is released
         // with the matching grown layout.
         unsafe { allocator.dealloc(grown_ptr, grown_layout) };

rustfs/src/memory_observability.rs

@@ -17,7 +17,10 @@ use rustfs_io_metrics::{
     record_cpu_usage, record_memory_usage, record_process_memory_split,
 };
 use serde::Serialize;
+#[cfg(any(test, not(target_os = "windows")))]
 use serde_json::Value;
+#[cfg(not(target_os = "windows"))]
+use std::ffi::CStr;
 use std::path::Path;
 use std::sync::{Arc, Mutex, OnceLock};
 use std::time::Duration;
@@ -228,18 +231,7 @@ fn read_cgroup_memory_snapshot() -> Option<CgroupMemorySnapshot> {
     read_cgroup_v2().or_else(read_cgroup_v1)
 }
 
-fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
-    let json = rustfs_mimalloc::MiMalloc::stats_json();
-    if json.is_empty() {
-        return None;
-    }
-    let observation = parse_mimalloc_stats_json(&json)?;
-    Some(AllocatorMemorySnapshot {
-        backend: crate::allocator_reclaim::allocator_backend(),
-        observation,
-    })
-}
-
+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_value(value: &Value) -> Option<u64> {
     match value {
         Value::Number(number) => number
@@ -250,6 +242,7 @@ fn numeric_json_value(value: &Value) -> Option<u64> {
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
     match value {
         Value::Object(fields) => fields
@@ -261,6 +254,7 @@ fn numeric_json_field(value: &Value, field: &str) -> Option<u64> {
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64> {
     match value {
         Value::Object(fields) => {
@@ -277,10 +271,12 @@ fn mimalloc_stat_field(value: &Value, metric: &str, field: &str) -> Option<u64>
     }
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_current(value: &Value, metric: &str) -> Option<u64> {
     mimalloc_stat_field(value, metric, "current")
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64> {
     metrics
         .iter()
@@ -289,6 +285,7 @@ fn mimalloc_stat_sum(value: &Value, metrics: &[&str], field: &str) -> Option<u64
         .filter(|value| *value > 0)
 }
 
+#[cfg(any(test, not(target_os = "windows")))]
 fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservation> {
     let value = serde_json::from_str::<Value>(stats_json).ok()?;
     let malloc_metrics = ["malloc_normal", "malloc_huge"];
@@ -315,6 +312,33 @@ fn parse_mimalloc_stats_json(stats_json: &str) -> Option<AllocatorMemoryObservat
     }
 }
 
+#[cfg(not(target_os = "windows"))]
+#[allow(unsafe_code)]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    // SAFETY: `mi_stats_get_json` returns a null-terminated JSON buffer owned by
+    // mimalloc when called with a null input buffer. The mimalloc API requires
+    // freeing that buffer with `mi_free`; parsing finishes before the buffer is freed.
+    let observation = unsafe {
+        let stats_ptr = libmimalloc_sys::mi_stats_get_json(0, std::ptr::null_mut());
+        if stats_ptr.is_null() {
+            return None;
+        }
+
+        let observation = CStr::from_ptr(stats_ptr).to_str().ok().and_then(parse_mimalloc_stats_json);
+        libmimalloc_sys::mi_free(stats_ptr.cast());
+        observation?
+    };
+    Some(AllocatorMemorySnapshot {
+        backend: crate::allocator_reclaim::allocator_backend(),
+        observation,
+    })
+}
+
+#[cfg(target_os = "windows")]
+fn read_allocator_memory_snapshot() -> Option<AllocatorMemorySnapshot> {
+    None
+}
+
 fn configured_memory_observability_interval_secs() -> u64 {
     rustfs_utils::get_env_u64(ENV_MEMORY_OBSERVABILITY_INTERVAL_SECS, DEFAULT_MEMORY_OBSERVABILITY_INTERVAL_SECS).max(1)
 }
@@ -542,13 +566,6 @@ mod tests {
         assert_eq!(parse_mimalloc_stats_json(r#"{ "allocator": "unknown" }"#), None);
     }
 
-    #[test]
-    fn read_allocator_memory_snapshot_uses_mimalloc_stats_json() {
-        let snapshot = super::read_allocator_memory_snapshot();
-        #[cfg(not(target_os = "windows"))]
-        assert!(snapshot.is_some(), "allocator snapshot should be available on non-Windows");
-    }
-
     #[test]
     fn memory_observability_snapshot_reports_disabled_when_metrics_are_disabled() {
         let snapshot = build_memory_observability_status_snapshot(false, 15, false);

rustfs/src/storage/options.rs

@@ -554,11 +554,10 @@ fn apply_replication_timestamps_from_headers(headers: &HeaderMap<HeaderValue>, o
 
     // Persist into the internal metadata keys so a later outbound replication
     // pass (replication_target_boundary) reads the source's modification
-    // times instead of falling back to mod_time. Receiver-side LWW happens at
-    // the set layer under the object write lock
-    // (ecstore set_disk::ops::object::merge_replication_metadata_lww,
-    // rustfs/backlog#1953): a category whose stored timestamp is newer than
-    // the inbound one keeps the local values.
+    // times instead of falling back to mod_time.
+    // TODO(P1-6): receiver-side LWW is still missing — when the stored
+    // per-category timestamp is newer than the inbound one, the existing
+    // tags/retention/legal-hold should win instead of being overwritten.
     for (timestamp, suffix) in [
         (opts.replication_tagging_timestamp, SUFFIX_TAGGING_TIMESTAMP),
         (opts.replication_retention_timestamp, SUFFIX_OBJECTLOCK_RETENTION_TIMESTAMP),