Track upstream #5140: post-review fails with 422 for out-of-hunk findings

Summary

Track the RHDH impact of fullsend-ai/agents#430.

Fullsend review runs on redhat-developer/rhdh-plugin-export-overlays#2778 complete the agent analysis successfully and update the sticky review comment, but fail when fullsend post-review submits the formal GitHub review. An out-of-hunk finding is converted to a file-level comment and remains part of the batched review comments; GitHub rejects the review with 422 Unprocessable Entity.

Evidence

Latest reproduction:

• PR: https://github.com/redhat-developer/rhdh-plugin-export-overlays/pull/2778
• Run: https://github.com/redhat-developer/rhdh-plugin-export-overlays/actions/runs/29501476073
• Fullsend version: v0.30.0
• Reviewed PR head: c501c8ae5f2da666c6872a5ba47e78cb931c5480

Relevant log:

1 finding(s) posted as file-level comment(s) (line outside diff hunk)
• Submitting APPROVE review
  Attaching 2 inline comment(s)
Error: submitting review: create pull request review on fullsend-ai/fullsend#2778: github api: 422 Unprocessable Entity

The out-of-hunk finding targets run-e2e.sh:33; the current PR diff for that file starts around line 63. The other finding targets a changed line in workspaces/topology/e2e-tests/tests/specs/topology.spec.ts.

This is the same failure mode described in upstream issue fullsend-ai/agents#430, including the same PR and an earlier run. The newer run shows that the problem also affects an APPROVE verdict on Fullsend v0.30.0.

Impact

• The agent's analysis succeeds.
• The sticky review comment is updated.
• No formal GitHub approval/review is created.
• The workflow and Fullsend status comment report failure, which is misleading and can block adoption or automation based on the review conclusion.

Tracking

• Follow upstream resolution in fullsend-ai/agents#430.
• Verify the fix is present in the Fullsend version used by the RHDH reusable workflow.
• Rerun review on an RHDH PR containing an out-of-hunk finding.
• Confirm the formal review is posted and the workflow completes successfully.

Interim workaround

Keep out-of-hunk findings in the sticky review body, but omit them from the batched formal-review comments until upstream posting supports them safely.

Start with upstream fullsend-ai/agents#430 and the RHDH reusable workflow path, then reproduce the failure using the cited PR, workflow run, Fullsend v0.30.0, and the out-of-hunk finding in run-e2e.sh:33. Done means verifying the upstream fix is available in the workflow's Fullsend version and rerunning a review to confirm the formal approval posts and the workflow succeeds.