[lesnik512]

Closes #319.

An unhealthy health check now answers 503 Service Unavailable instead of 500, on FastAPI, Litestar, FastMCP and FastStream. The route is unhealthy when a probe returns False or raises, and on FastStream also when a broker ping fails. The JSON body is unchanged.

A 500 says the handler failed. A 503 says the service can't serve right now, so alerting can tell a dependency the probe found down apart from a crashing handler. Both are 5xx, so a rule on all 5xx still fires. Kubernetes treats any status from 200 to 399 as success, so probes behave the same as before.

Changes

• FastAPI, Litestar and FastMCP use HTTP_503_SERVICE_UNAVAILABLE from fastapi.status, litestar.status_codes and starlette.status. FastStream's literal changes from 500 to 503.
• The "Health checks" section of configuration.md gives the new status.

Tests

Written first, as the issue lists. The two failing cases in HEALTH_PROBE_CASES and the three FastStream assertions now expect 503. On main, 11 cases failed across all four frameworks:

• the one-returns-false and one-raises cases of each framework's test_*_health_check_runs_probes
• test_faststream_bootstrap_health_check_wo_broker
• test_faststream_health_check_fails_when_second_broker_is_down
• test_faststream_health_check_fails_on_broker_ping_despite_passing_probe

The full suite passes, 557 tests at 100% coverage (just test-ci). just lint (including ty), just adr-check and just docs-build pass too. No floor changes.

Breaking changes

Monitoring or load-balancer rules that match a 500 from the health route stop matching. Match 503, or any 5xx.