**Keep unread replication inventory distinct from absent jobs**
Validate complete typed replication snapshots before using their counts
or job health. Malformed JSON, lists, flags and ambiguous identities use
the existing safe error and retry instead of hiding the view, dropping
rows or coercing health. Preserve scoped job IDs, accepted empty reads
and existing poll/access semantics.
**Verify replication inventory failures through the production app**
Record parent-bound desktop Chromium and narrow WebKit consumer proof,
native input, safe error/retry, scoped job state and
complete/empty/denied observation semantics. Lane evidence retains
source VM receipts, raw browser results, inspected captures and the two
corrected driver attempts.
**Keep replication registry entries in canonical order**
Sort the new owned-file and exact-proof lists with registry_audit's
case-folded lexical order. Preserve every member, proof policy, frontend
byte and existing parent-bound browser receipt; repair the recorded
preflight rejection without changing product behaviour.
---
Fleet candidate `20261030T090451Z-web-product` from the `web-product`
lane. Reviewed tip `c07cbf886c47f3151ad79d0239f75d94a3a3fc75` on `main`.
[How Pulse handles
triage](https://github.com/rcourtman/Pulse/blob/main/docs/AI_TRANSPARENCY.md)
pulse-triage[bot]
1d8d84f7af