2026-06-24

Queue Observability Snapshot — 2026-06-24

Queue Depth

Queue	Count
done	446
active	7
blocked	6
queued	1

Throughput (last 24h)

• Tasks completed: 67 (healthy throughput)

DLQ Signals (last 24h, scanned files in done/)

Signal	Count	Status
API connection failures	12	⚠️ elevated
HTTP 500	0	✅ normal
Timeouts	0	✅ normal
Tool missing (exit 127)	0	✅ normal
Unknown failure tag	0	✅ clean (classifier working)
CrashLoopBackOff	1	low
OOMKilled	0	✅ normal
Assertion failed	0	✅ normal
CI test timeout	0	✅ normal

Stuck Tasks (>4h idle in active/)

• c1790169.md — 13.0h idle (decomposed parent not auto-archived)
• fix-test-assertion-failed-20260623.md — 11.8h idle
• fix-test-ci-timeout-20260623.md — 9.6h idle
• crashloop-hermes-qdrant-2026-06-22.md — 5.8h idle

Blocked Queue (oldest first)

File	Age	Priority
calliope-pvc-io-error-2026-06-18.md	6.0d	⚠️ >4d, needs pvs review
add-gpu-node-health-check-fallback-before-ocr-fa.md	1.9d	—
b4ed689e-diagnose-hermes-agent-crashloopbackoff-27-restarts.md	1.4d	—
investigate-stuck-tasks-2026-06-23.md	0.9d	—
fix-infra-crashloopbackoff-20260623.md	0.7d	—

Metrics File

• Last modified: 3.3h ago (fresh)
• Total lines: 540 (cumulative)

Anomalies

1. API connection failures elevated — 12 in 24h vs baseline ~3/day. Not a spike per se, but above normal. Likely transient network blips or API gateway retries. Monitor next run.
2. 4 stuck tasks in active queue — idle >4h without heartbeat. c1790169 is a decomposed parent that should have been archived to done/ per dispatcher convention. The other 3 are fix-it cards that may be waiting on human follow-up or external blockers.

Action Items

• Low priority: Move c1790169.md from active→done (decomposed, children exist)
• Monitor: API connection failure count next session — if >20, investigate infra
• Stale blocked item: calliope-pvc-io-error at 6d old — flag for pvs review per skill convention