handleHealth returns static ok (backs the App Runner probe); no collected/rejected/bot-blocked counters; no ingest-drop alarm — the CORS bug shipped twice with nothing to detect it. Fix: health check does a cheap getItem; per-site hourly counters in-table + ops endpoint + day-over-day drop alert on the jobs runtime; leveled JSON logger replacing 197 unstructured console.* calls.
From the production-readiness audit (docs/research/production-readiness.md, eacf996). Verified with live probes where stated.