[aw-failures] Fix: Smoke Claude safe_outputs fails when resolve_pull_request_review_thread targets already-resolved thread

[github-actions]

Problem Statement

The Smoke Claude smoke test run §25181816514 (2026-04-30 18:14 UTC) failed at the safe_outputs step due to a GraphQL error on resolve_pull_request_review_thread. All 19 smoke sub-tests completed (18 pass, 1 skipped), but the safe_outputs stage failed, marking the entire run as a failure.

Error Signature

Processing message 8/11: resolve_pull_request_review_thread
Failed to resolve review thread: Request failed due to following response errors:
Message 8 (resolve_pull_request_review_thread) failed: Request failed due to following response errors:
1 safe output(s) failed:
  - resolve_pull_request_review_thread: Request failed due to following response errors:

Thread ID attempted: PRRT_kwDOPc1QR85-0kYa (on PR #29360)

Root Cause

The Smoke Claude smoke test:

1. Calls submit_pull_request_review (creates a new review on PR fix: extend detectFirewallAuditArtifacts to find audit files in non-flattened agent artifact #29360)
2. Then calls resolve_pull_request_review_thread targeting a pre-existing thread ID (PRRT_kwDOPc1QR85-0kYa)
3. Queues this resolve in safeoutputs.jsonl

The thread referenced in step 2 may already be in a resolved state from a previous Smoke Claude run. The safe_outputs handler then fails when trying to re-resolve an already-resolved thread via the GraphQL API.

Impact

• Smoke Claude test incorrectly shows as FAIL despite all sub-tests passing
• 1 safe output failure triggers full safe_outputs step failure
• Creates a misleading signal that the Claude engine is unhealthy when it is not

Metrics

Metric	Value
Agent status	success (9.3m, 42 turns)
Smoke sub-tests passed	18/19 (test 19 skipped)
safe_outputs step	FAILURE (22s, 1/11 items failed)
Tokens	1.23M
Cost	~/bin/bash.98

Proposed Remediation

Option A (Recommended): Resolve the thread created in the same run

• The smoke test should capture the thread ID of the review it just created (from the submit_pull_request_review response) and resolve THAT thread, not a hardcoded pre-existing thread ID

Option B: Use continue-on-error for resolve_pull_request_review_thread in safe_outputs

• Configure the safe_outputs handler to treat resolve_thread failures as warnings rather than errors, since the thread being already-resolved is a harmless state

Success Criteria

Smoke Claude runs to full success. resolve_pull_request_review_thread resolves a thread from the same run and does not fail when threads are pre-resolved.

Parent: #29232

References:

• §25181816514 (failing run)

Generated by [aw] Failure Investigator (6h) · ● 397.8K · ◷

• expires on May 7, 2026, 7:24 PM UTC

[github-actions]

🍪 Issue Monster selected this for Copilot

I've identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · ● 18.3K · ◷