The registry-image route needed a working CI runner + owner-namespaced registry push; not worth it for a cosmetic image_sha. Coolify builds from the connected repo as before.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Traefik runs on the `coolify` shared network. Without the gateway also
being on it, Traefik logs "Could not find network 'coolify' for
container, defaulting to first available" and picks an IP on the
project-local network it can't reach → intermittent 504 Gateway Timeout
(30s) in the browser.
Hot-patched the running container via `docker network connect coolify`;
this commit makes the fix permanent so the next Coolify redeploy doesn't
reintroduce the regression.
Worker and cron don't need it — they don't serve external HTTP.
`timescale_db` alias resolves. Without this the new app's PG pool
times out — DB only has the alias on its own project network, not on
`coolify` shared.
Three services (gateway/worker/cron) from one build, env-driven, no
explicit Traefik labels (Coolify generates them from the Domain
setting). Replaces the manual `docker run` containers that have been
running since the first VPS deploy.
