ultrareview: failed reviews count against free tier (server crashes before findings)
Resolved 💬 1 comment Opened May 4, 2026 by vonaxai Closed Jun 1, 2026
Summary
Both my ultrareview attempts on PR #8 (vonaxai/vonaxai-system, private repo) failed with "Review crashed before producing findings" — but BOTH counted against my 3 free-tier reviews. I now have 1 of 3 left despite getting zero useful output.
Steps to reproduce
cd <repo>claude ultrareview 8(PR with ~38 commits, several hundred file changes — recovery branch)- Output:
"Free ultrareview 1 of 3"— slot consumed - After ~5 min:
"Review failed: Review crashed before producing findings" - Retry:
claude ultrareview 8→ again"Free ultrareview 2 of 3", again same crash
Sessions (server logs should be available there)
- https://claude.ai/code/session_012MUXxQBjsB8oU68oX29b3b
- (second session URL not captured in stdout — please look up by user)
Expected
- If the cloud reviewer can't handle the PR (size, complexity, etc.), return a clear error like "PR too large for cloud review" and don't decrement the free-tier count.
- If it crashes server-side, don't decrement the free-tier count — failed reviews shouldn't be billed.
Actual
- Crash happened server-side (no output produced)
- Counter went
1/3 → 2/3anyway
Environment
- Claude Code:
2.1.126 - Platform:
darwin - Local repo: 38-commit PR against main, primarily docs/configs/scripts/landing-pages
Ask
Refund the 2 consumed free reviews + fix the counting so server-side crashes don't decrement the quota.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗