Health contract
SELECT 1 proves a limited database round trip. It does not validate schema migrations, data quality, import recency, connection-pool headroom or query performance under load. Treat the health measurement as a connectivity signal and keep those application-level concerns distinct.
Independent frontend and backend checks
Separate targets make it possible to identify a reachable frontend with an unhealthy backend. The probe records TLS failure or missing HTTP response as unsuccessful, never as a healthy fallback. Timing ends at response headers and is not a browser rendering benchmark.
Shared probe dependency
Monitoring runs on the local server. Internet or host failures can remove evidence for several sites simultaneously. Historical storage remains in Grafana Cloud, but an absent sample cannot establish that the Fantasy service itself is down. Consult last-check freshness and the local host dashboard.
Further instrumentation
Importer age, request errors and process resource usage need dedicated application or host signals before they can be shown accurately. This view intentionally contains no fabricated values for those measurements.