[rmcdaniel]

Startup diagnostics silently omitted oversized/streamed/invalid response metadata and records evicted from bounded buffers. Existing worker-history requests were also counted as submissions. The observer now records explicit metadata states and omission counts, marks truncated buffers/milestone events, and classifies existing history requests separately without consuming streams or issuing probes.

This is the first small observer-source slice for #123. The flags describe retained evidence only; they establish no task/phase verdict, historical root cause or installed-runtime acceptance. Original source pins/failure distinctions, case budgets, Worker/Client behavior and the pending SQL/API comparator are preserved. No SDK product source or workflow changes.

Validation:30 focused source tests pass; Ruff and mypy42 source files pass. Inert baseline controls reproduce the old history-submission classification and missing loss flags; no network is used locally. Normal existing source CI is required. This ordinary source PR carries no package publication, promotion, historical execution or #184 authority.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[rmcdaniel]

Owner source review at exact2968624eb624ad4efb447771f68ab90822b27672/treea1b0a1a521b79e59fda273cdf36bab883f3aae36, base7c8e3821: all20 checks are terminal SUCCESS, including CI42083790411, all six supported Python versions, integration, target qualification, Sonar, CodeQL and public boundary. Thirty local focused tests/Ruff/mypy pass; baseline controls reproduce the former history-submission and silent-loss behavior.

Reviewed the insertion/change scope and report limits: existing request/response semantics, stream non-consumption, redaction and budgets remain; no observer probe, SDK product source or workflow is changed. Explicit completeness flags cover retained metadata/buffers only, with SQL/API/full phase linkage and historical execution still pending on #123.

Automatic publication followers checked before integration: no workflow_run publisher exists; PyPI publication requires a tag or explicit protected manual request. Docs push paths exclude these test files, and stable-tag/deployment logic is untouched. Ordinary source integration carries no artifact publication/promotion or historical/frozen acceptance. Proceeding through the normal owner merge path with exact-head binding, then actual-main verification/branch cleanup.

[rmcdaniel]

Ordinary source integration and actual-main follow-through are complete for #278. Merge 78ff7a1 has exact reviewed tree a1b0a1a521b79e59fda273cdf36bab883f3aae36, ordered parents[7c8e382,2968624]. All18 actual-main checks and all four triggered workflows pass: CI42085629324, Sonar42085629072, CodeQL42085628491 and boundary42085629168. No skipped or failed check is treated as green.

The test-only paths triggered no docs or publication workflow. Fresh five-page bytes match restored published2.14.1 artifact13380779871, with all five queued/running/waiting/pending/requested docs-writer counts zero. No docs dispatch, release/tag, image/package promotion or new publication occurred.

The exact merged remote head2968624 was deleted and independently verified absent. The clean shared main is fast-forwarded to78ff7a18, and the isolated merged worktree/local branch were removed. #123 remains open: metadata/retention flags and existing history-route classification are delivered as source; SQL/API snapshots, full phase linkage, historical execution and the original unexplained failures remain separate. sdk-python owns the next source slices and the independent #184 composed-custody successor, whose operating gates remain held.