Health probe used by Lightsail + uptime monitors.

Reports DB connectivity, disk-space utilisation against the data partition, and Redis ping. Returns 503 when the database can't be reached; 200 in all other cases (including "warning" disk levels — the field-level details let monitors distinguish).

The HEAD twin (:func:health_check_head) delegates here so UptimeRobot / Pingdom / Datadog Synthetics — which default to HEAD — also get a 200 instead of a 405.

Response is cached in-process for :data:_HEALTH_CACHE_TTL_S seconds — see the module-level note for the rationale.

Detail-field exposure is gated (THECOLONYC-431): the full checks internals go only to an internal caller (loopback socket peer — the in-container operator / docker healthcheck); a request arriving via nginx gets the redacted view (:func:_redact_health_checks). The top-level status and status code are identical for both. Cache-Control: no-store keeps any shared cache from holding the (potentially privileged) body.