review-runs-tracking: 2026-05

[dormouse-bot]

Monthly tracking issue for below-threshold findings. Each run appends findings as a comment. Future runs read these to build cumulative evidence.

Do not close manually — a new issue is created each month, and prior months are closed automatically.

[dormouse-bot]

Run 26328198983 — 2026-05-23 (review-runs)

tend-review spends $0.56–0.77 per no-op self-review of the bot's own PRs

• Evidence level: Low/Medium
• Occurrences this run: 3 (docs: sync tutorial/AGENTS pointers after playground route split #88, docs: sync deploy.md with current release workflow #89, chore: update tend workflows #90)
• Run IDs: 26326753836, 26326777937, 26326814709
• Workflow: https://github.com/diffplug/dormouse/actions/runs/26328198983
• Detail: The nightly run opened PRs docs: sync tutorial/AGENTS pointers after playground route split #88–90; tend-review then fired on each. The bot correctly produced no review/comment (self-review guard worked as intended), but each run still read 450–750K cache tokens and ran several turns before exiting (~$0.56–0.77 each, ~$2 total). Behaviorally correct; flagged only as a possible efficiency opportunity (e.g. an earlier author==bot short-circuit). Structural trigger pattern, but not a defect — recording for accumulation, not acting.

tend-install-test "Verify generator output matches committed files" failed once, then self-healed

• Evidence level: Low
• Occurrences this run: 1 (self-resolved)
• Run ID: 26300228733
• Workflow: https://github.com/diffplug/dormouse/actions/runs/26328198983
• Detail: Single install-test failure at 16:41 on the "Verify generator output matches committed files" step (committed tend workflows drifted from generator output). The 6 subsequent install-test runs passed, and the nightly opened PR chore: update tend workflows #90 ("chore: update tend workflows") which addresses the drift. This is a generator-verification test, not a behavioral Claude run — no action needed; noting for pattern tracking in case drift recurs.