fix(heartbeat): stop welding a stale reviewState onto a new same-PR review instance (BLO-22229)

[allyblockcast]

Thinking Path

• Paperclip coordinates AI agents, GitHub events, and issue work through heartbeat wakeups.
• GitHub PR review wakeups carry review state, author, body, comment ids, and structured wake context into author-facing directives.
• Active heartbeat runs can coalesce multiple wakes for the same issue/PR, so stale review fields or stale comment ids can be inherited if replacement is partial.
• BLO-22229 found a same-PR review-instance bug: a comment-shaped Ally review could inherit an older formal approval state.
• Review feedback also showed the reverse sequence could keep superseded feedback comments in wakeCommentIds after a later formal review took ownership.
• This pull request makes review-instance wakes replace review content as a unit and retire superseded review-feedback comment ids while preserving unrelated human comments.
• The benefit is that author wakes no longer mix an approval state with stale Critical/Important findings from a different review instance.

Linked Issues or Issue Description

Refs: https://paperclip.blockcast.net/BLO/issues/BLO-22229
Refs: https://paperclip.blockcast.net/BLO/issues/BLO-19118

Bug: Same-PR review wake coalescing could compose context from two different review instances. A later comment-shaped review lacked its own formal reviewState and inherited an earlier approval; a later formal review could also keep an older actionable review-feedback comment in wakeCommentIds, causing the rebuilt paperclipWake.comments payload to reload superseded findings.

What Changed

• Treat github_pr_review_submitted and github_pr_review_feedback wakes as fresh review instances that own review-content fields outright.
• Clear review-content fields the incoming review instance does not supply instead of inheriting them from an earlier same-PR review.
• Filter superseded github_pr_review_feedback comment ids out of the existing wake-comment set when a same-PR review instance replaces the prior review context.
• Preserve unrelated human comment ids during that filtering so ordinary comments still remain available to the next wake payload.
• Added regression coverage for stale approval-state inheritance, formal-review replacement, and feedback-to-approval comment retirement.

Verification

• PATH=/opt/homebrew/bin:$PATH pnpm install --frozen-lockfile
• PATH=/opt/homebrew/bin:$PATH pnpm exec vitest run server/src/__tests__/heartbeat-context-summary.test.ts server/src/__tests__/heartbeat-workspace-session.test.ts --no-file-parallelism --maxWorkers=1 (283 passed)
• PATH=/opt/homebrew/bin:$PATH pnpm --filter @paperclipai/server typecheck
• PATH=/opt/homebrew/bin:$PATH pnpm exec vitest run server/src/__tests__/workspace-runtime.test.ts --testNamePattern "provisions worktree-local pnpm node_modules instead of reusing base-repo links" --no-file-parallelism --maxWorkers=1 (2 passed, 116 skipped)
• PATH=/opt/homebrew/bin:$PATH pnpm exec vitest run server/src/__tests__/heartbeat-queued-backlog-convergence.test.ts --no-file-parallelism --maxWorkers=1 (11 skipped locally)

Risks

Low to moderate risk. The change is scoped to heartbeat coalescing for GitHub review-instance wakes. The main behavior shift is intentional: a new review instance no longer inherits stale review-content fields or stale review-feedback comments from a prior review instance. Non-review wakes still inherit existing review context.

Model Used

Codex GPT-5 coding agent with shell, GitHub CLI, and local test execution.

Checklist

• I have included a thinking path that traces from project context to this change
• I have specified the model used (with version and capability details)
• I have checked ROADMAP.md and confirmed this PR does not duplicate planned core work
• I have searched GitHub for duplicate or related PRs and linked them above
• I have either (a) linked existing issues with Fixes: # / Closes # / Refs # OR (b) described the issue in-PR following the relevant issue template
• I have run tests locally and they pass
• I have added or updated tests where applicable
• If this change affects the UI, I have included before/after screenshots
• I have updated relevant documentation to reflect my changes
• I have considered and documented any risks above
• All Paperclip CI gates are green
• Greptile is 5/5 with no open P2s, recommendations, or follow-ups
• I will address all Greptile and reviewer comments before requesting merge

[allyblockcast]

🔗 Paperclip issue: BLO-22229
🔗 Paperclip issue: BLO-19118

[allyblockcast]

🔗 Paperclip issue: BLO-22229
🔗 Paperclip issue: BLO-19118

[allyblockcast]

Hey @allyblockcast[bot]! Before this PR can be reviewed, a few things need attention:

Missing or incomplete:

• Missing section: ## Thinking Path
• Missing section: ## What Changed
• Missing section: ## Verification
• Missing section: ## Risks
• Missing section: ## Model Used
• Add the dedup-search checkbox to your PR description and check it once you have searched the GitHub PR list for similar PRs. See the PR template at .github/PULL_REQUEST_TEMPLATE.md and CONTRIBUTING.md → "Before You Start: Search First".

Once updated, push a new commit and these checks will re-run automatically.

— commitperclip

[allyblockcast]

Ally — Consolidated PR Review

Lenses: pr-review-toolkit (code, tests, comments, errors, types) + gstack/review + native-codex.
Reviewed head: a76d7b5

Important Issues (1)

• [gstack/review] server/src/services/heartbeat.ts:6553 — Replacing the review fields does not retire the prior review's feedback comment from wakeCommentIds. After this loop clears githubReviewFeedbackCommentId, line 6567 still merges every comment ID from existing; buildPaperclipWakePayload then reloads those comments. In the reverse sequence covered by the incident class (comment-shaped actionable feedback followed by a formal APPROVED review), the author wake can therefore carry the new approval state while paperclipWake.comments still contains the superseded Critical/Important findings.
  • Remove the superseded review-feedback comment IDs from the existing comment set when a new review instance takes ownership, while preserving unrelated human comments. Add a feedback-to-approval regression test whose existing fixture includes githubReviewFeedbackCommentId, wakeCommentIds, and paperclipWake, and assert that only non-review comments survive.

Strengths

• The new ownership rule correctly prevents a missing incoming githubPrReviewState or body from inheriting the previous review instance's value.
• The tests cover both formal-review-to-comment-feedback and formal-review-to-formal-review field replacement.

Recommended Action

1. Fix the Important issue before merge and rerun the focused heartbeat tests.
2. The PR is authored by app/allyblockcast; the exact head must be reopened under an independent author before an Ally App approval is possible.

[kkroo]

Superseded by #1064, which carries the same BLO-22229 fix plus the follow-up for this review feedback from a human-authored branch so Ally can review it.