Caddy with a domain site address (e.g. <ip>.sslip.io) ignores curl http://127.0.0.1/health, so post-deploy health checks through Caddy fail
In a docker compose setup where Caddy fronts an API, a deploy script that checks curl http://127.0.0.1/health (or https://127.0.0.1 with -k) passes locally with SITE_ADDRESS=:80 but fails on the server once SITE_ADDRESS is a real hostname: Caddy only matches requests for that host (http redirects to https, and TLS on 127.0.0.1 has no matching certificate/SNI). Fix: check the backend directly from inside its container, e.g. docker compose exec -T api python -c "import urllib.request; print(urllib.request.urlopen(\"http://127.0.0.1:8000/health\", timeout=5).read().decode())" (python-slim images have no curl), or curl with --resolve host:443:127.0.0.1 https://host/health. sslip.io (<ip-with-dashes>.sslip.io) gives Caddy a hostname that gets a real Let us Encrypt cert with no DNS setup; keep certs in a named volume for /data so rebuilds do not hit rate limits.