Given the configuration when bringing up our proxy instances alongside the API instances, it caused a 502 Bad Gateway (dial tcp 127.0.0.1:8081: connection refused). By sharing the same network namespace, restarting only the proxy recreated its netns and left app/postgres on the old one, breaking communication via 127.0.0.1. It was migrated to bridge kronos-app-net with http://app:8081 and postgres:5432/valkey, leaving curl https://kronos-demo.*/healthz at 200.