Locate telemetry loss with per-boundary flow evidence, queue age and a known canary while accounting for retries, filtering and temporality.
Combine seasonal traffic diagnostics with SLO burn alerts and explicit low-volume policies without allowing baselines to learn an outage.
Preserve regional failure boundaries in SLIs and dashboards, detect missing regions explicitly, and avoid global aggregates that conceal outages.
Alert on queue age, deadlines and stalled progress while distinguishing normal batch accumulation from idle or failed consumers.
