Enhancement: Add PR state validation and auto-recovery for cross-repository forks

[konard]

Background

Based on the case study in #1214, we identified that cross-repository fork PRs can be unexpectedly closed by GitHub during push operations. This is a GitHub platform issue, but we can implement mitigations in hive-mind.

Related

• Case study: Pull Request was closed #1214 and PR Case study: PR unexpectedly closed during AI work session #1215
• Full analysis: docs/case-studies/issue-1214/CASE-STUDY.md

Proposed Enhancements

1. PR State Validation After Push

Add validation after every push to detect unexpected PR closures:

async function validatePRState(owner, repo, prNumber) {
  const state = await gh(`api repos/${owner}/${repo}/pulls/${prNumber} --jq .state`);
  if (state !== 'open') {
    throw new Error(`PR #${prNumber} is ${state}, expected 'open'`);
  }
}

// After push
await sleep(5000); // Wait for GitHub to sync
await validatePRState(owner, repo, prNumber);

2. Auto-Reopen Mechanism

Implement automatic PR reopening when an unexpected close is detected:

async function ensurePROpen(owner, repo, prNumber) {
  const pr = await getPRDetails(owner, repo, prNumber);
  if (pr.state === 'closed' && !pr.merged) {
    console.log(`PR #${prNumber} was unexpectedly closed, attempting to reopen...`);
    await gh(`pr reopen ${prNumber} --repo ${owner}/${repo}`);
    await notifyOwner(owner, repo, prNumber, 'PR was automatically reopened after unexpected closure');
  }
}

3. Periodic State Monitoring

During long-running AI sessions, periodically check PR state (every 2-3 minutes):

const stateCheckInterval = setInterval(async () => {
  await validatePRState(owner, repo, prNumber);
}, 120000); // Every 2 minutes

4. Cross-Repository Awareness

Add special handling for isCrossRepository: true PRs:

• Extra delay after push operations (5-10 seconds) for GitHub sync
• More frequent state validation
• Automatic reopen attempts with backoff

Acceptance Criteria

• After every push, validate PR state is still 'open'
• If PR is unexpectedly closed, attempt automatic reopen
• If reopen fails, notify via PR comment and raise clear error
• Add logging for PR state changes
• Add telemetry to track frequency of unexpected closures

Priority

Medium - The issue is rare but disruptive when it occurs.

---

Generated from case study analysis