Shaped like Statuspage v2, so a dashboard or bot that already watches a status page works against Foxl with only the host changed. Every response is public, CORS-open, and served from a precomputed snapshot, so polling it costs us one indexed row read at most.

monitoring reports the freshness of the monitoring itself (stale, age_ms, measured_at). Alarm on it: a status page whose own cron has died otherwise reports green forever. unknown can appear as an indicator and as a component status, meaning the probe could not establish either state. Code that treats anything other than operational as not-fine handles it correctly.

Components are probed every 2 to 15 minutes depending on cost, so polling faster than once a minute tells you nothing new. cache-control says how long each response is good for.