feat(observability): report event-loop stalls and runtime pressure #1132

Outcome

The monitoring endpoint exposes health metrics; recent upstream diagnostics propose an event-loop stall sampler. Audit current lag sampling before adding only the missing counter/report.

Evidence and source state

DurinDoor baseline: 93f6de64c69b17d52f790d6f14d09a74cffc7fa5.

• src/app/api/monitoring/route.js
• omniroute #15888: fix(monitoring): log event-loop stalls with counter on health | OPEN
• omniroute #15978: fix(monitoring): warn once on unusable stall sampler reading | OPEN

Implementation subtasks and acceptance criteria

• Check existing latency probes and sampling overhead.
• Add bounded diagnostic events and correlate pressure with admission failures.
• Handle unusable sampler readings without repeated log spam.

Dependencies

No hard predecessor. Complete source review and reproduction/discovery before implementation.

Parent epic: #1063. Roadmap: #1056. Priority P2. Stage 3.

Delivery contract

Planning ticket only. No implementation, runtime test, live-provider call, deployment or stored-data mutation was performed in this audit. Before a port, confirm a merged final upstream diff or a reachable source commit, compare local equivalents/tests, and preserve local auth/routing/quota/database contracts. Open upstream work remains a review candidate. Use port(omniroute): title and target a focused PR at bloodf/durindoor:main. Behavior changes require a focused test plus docs. Run npm run lint and cd tests && npm run test:ci; never add baseline failures. Pure docs/CI edits may skip runtime tests. Regenerate registry/agent indexes with their scripts. Preserve stored credentials and unrelated work.

Implementation moved to #1129

The complete original scope and acceptance criteria are preserved in #1129. Follow that canonical ticket for implementation and verification. This consolidation does not mean the work shipped. Original discussion and source links remain here.