2026-05-10.md

🩺 Daily 🟡 9 🟢 5

Daily Health Check — 2026-05-10

[🟡] Codex timeout/fetch-error cluster u zadnja 24h — journald pokazuje više openai-codex timeout/fetch failover događaja između 12:28-15:44 (network connection error, active-memory ... status=timeout, embedded run failover decision), ali bez refresh_token_reused, usage-limit ni Codex app-server required runtime mismatch signala. Ovo izgleda kao prolazna dostupnost/latency degradacija, ne OAuth problem (severity: medium)
[🟡] daily_logs pressure se pogoršao novim oversized logom — memory-health-breakdown.py --json i dalje flagga daily_logs, a sada je i 2026-05-09.md oversized na 31.7 KB / 124 linije uz stare velike datoteke iz travnja. Problem je i dalje lokaliziran samo na daily_logs bucket, ne na cijeli memory root (severity: medium)

[Weekly Regression Suite i dalje otvoren incident od restarta] 🟡→🟡 IN_PROGRESS (openclaw cron list --json još pokazuje lastStatus: error i consecutiveErrors: 1, ali bez novog runa/faila u zadnjih 24h; ostali cronovi su OK.)
[Codex stale OAuth profile refresh greške] 🟡→🟢 RESOLVED (u zadnjih 24h nema refresh_token_reused ni Token refresh failed: 401 signala; novi codex problemi su timeout/fetch-error prirode.)
[daily_logs bucket pressure] 🟡→🟡 IN_PROGRESS (daily_logs ostaje warn, a lista oversized fajlova sad uključuje i novi 2026-05-09.md.)
[runtime_state (memory/.dreams) over threshold] 🟡→🟡 IN_PROGRESS (runtime_state je i dalje warn, sada ~588 KB vs 300 KB prag; dominantno short-term-recall.json 462 KB i events.jsonl 140 KB.)
[Config drift watch] 🟢→🟢 STABLE (na kanonskim AGENTS.md, TOOLS.md, DECISIONS.md, SOUL.md i openclaw.json putanjama nema ctime promjena u zadnjih 24h.)
[Known codex warmup Unknown model] 🟢→🟢 IGNORED (pgrep -af 'codex app-server' vraća aktivan proces, pa je warmup signal tretiran kao poznat/benigan i nije finding.)

2 new findings (0 critical, 0 high). Token cost: ~minimal.