[regression] Daily CI Regression Audit — 35048318840

[github-actions]

Summary

• CI failure rate (last 20 runs): 6/20 = 30% (also 4 cancelled runs)
• Consecutive failure streak: runs [safe-output-integrator] Add safe-output coverage for upload-code-coverage #55144–[aw-failures] [P0] copilot-sdk headless server SIGABRT before ready — Dead Code Removal Agent total run loss #55147 (4 in a row, 2026-09-15 19:58–20:25 UTC)
• Recommended action: 🔎 Investigate — failure rate exceeds the 10% healthy threshold and includes a real streak, though root causes appear to be a mix of one flaky/transient issue and one repo-hygiene (lock-file drift) issue rather than a single code regression.

Failure Details

Runs inspected (most recent 3 failures out of the last 20 completed CI runs):

Run #	Run ID	Conclusion	Failing job	Root cause
55155	35034674116	failure	Integration: CLI Logs Live Cache (TestLogsCachedJSONLLiveCaching)	Test assertion failure: "run 34855185106 should occur in exactly one JSONL cache shard" — logs-cache sharding logic wrote/matched a run ID in more than one JSONL cache shard file.
55147	35019492212	failure	Check workflow markdown/lock drift	A .lock.yml file was out of sync with its .md source (make recompile not run/committed before merge — CI enforces lock/markdown drift check).
55146	35018455341	failure	Integration Unauthenticated Add (Public Repo)	Transient network failure: go: google.golang.org/protobuf@v1.36.12: ... stream error: stream ID 693; INTERNAL_ERROR — Go module proxy connectivity blip, not code-related.

Top failure patterns identified:

1. Flaky/transient infra failure (1 of 3): Go module proxy (proxy.golang.org) HTTP/2 stream error during dependency fetch — classic network flakiness, no code change implicated.
2. Lock-file drift regression (1 of 3): a PR merged with an out-of-date generated .lock.yml, failing the "Check workflow markdown/lock drift" gate. Recurrent risk if make recompile is skipped before commit.
3. New/isolated test regression candidate (1 of 3): TestLogsCachedJSONLLiveCaching failed on a JSONL cache-sharding invariant ("run should occur in exactly one shard"). This did not recur in the other 2 sampled failed runs' job lists (jobs of the same name passed elsewhere), suggesting either a flaky race in the live-cache sharding test or a narrow regression tied to a specific commit — worth a follow-up look if it repeats.

No single pattern repeated across all 3 failures — these look like independent, uncorrelated failures (1 flaky network, 1 process/hygiene gap, 1 possible test flake/regression) rather than one systemic regression.

Open regression/bug issues (labels: bug, regression, top 5):

• [deep-report] push_repo_memory silently deletes repo-memory checkout while reporting false success #60773 — [deep-report] push_repo_memory silently deletes repo-memory checkout while reporting false success
• pushSignedCommits fallback pushes stale branch ref after rebasing detached HEAD #58910 — pushSignedCommits fallback pushes stale branch ref after rebasing detached HEAD
• [WIP] Smoke Project: work in progress #57272 — [WIP] Smoke Project: work in progress
• [Firewall Test] Escape test failed #57063 — [Firewall Test] Escape test failed

None of these open issues directly match the 3 failure patterns found above (no existing tracked issue for TestLogsCachedJSONLLiveCaching sharding or the lock-drift gate failure was found).

Generated by 🧭 Daily Regression Audit · copilot · auto · 51.5 AIC · ⌖ 6.6 AIC · ⊞ 7.1K · ◷

• expires on Sep 17, 2026, 6:40 PM UTC-08:00

[github-actions]

This issue is being closed as outdated. A newer issue has been created: #61495

View newer issue

---

This action was performed automatically by the Daily Regression Audit workflow.