Bug: Frontend redirects to `/unreachable` on backend cold-start instead of showing a loading state

[max-tet]

Is this urgent?

No

What parts are affected

Frontend

What is the server version

v0.7.9

What is the client version

v0.7.9

What platform are you using

Linux

What's the problem 🤔

Quick context first: I run KitchenOwl on Freeshard, a self-hosting platform where apps are put to sleep when idle and restarted on demand. So a cold start is totally normal here — you open the app and the containers spin up on the spot, with the web container coming up a few seconds before the backend is ready. That timing is what surfaces this.

Setup: split kitchenowl-web + kitchenowl-backend.

What happens: when I open the app while the backend is still starting, the frontend loads from the service worker cache and immediately redirects to /unreachable. So instead of a "starting up, hang on" screen I just hit the dead-end page and have to reload a few times until the backend is up.

Two flavours of the same thing:

1. Backend briefly unreachable during startup → hard redirect to /unreachable instead of a loading/retry state.
2. Right after a version bump, the cached frontend is older than the new backend's min_frontend_version, so I get "app version is not supported, please update the app" until I hard-refresh (Ctrl+F5) to drop the SW cache.

What I'd expect: when the health check fails transiently (connection refused / timeout), show a loading / "backend starting" state and retry for a bit before falling back to /unreachable. And a stale service-worker cache shouldn't hard-fail on the version check — ideally the SW updates and reloads itself.

tbh the redirect makes cold-starts look broken when they're really just slow. A short retry loop would smooth it over.

Happy to test a fix on my setup.

Share your configuration

networks:
    portal:
        external: true

services:
    kitchenowl:
        restart: always
        image: tombursch/kitchenowl-web:v0.7.9
        container_name: kitchenowl
        volumes:
        - "{{ fs.app_data }}/data:/data"
        environment:
        - BACK_URL=kitchenowl-backend:5000
        depends_on:
            kitchenowl-backend:
                condition: service_healthy
        networks:
        - portal
    kitchenowl-backend:
        restart: always
        image: tombursch/kitchenowl-backend:v0.7.9
        container_name: kitchenowl-backend
        environment:
          - JWT_SECRET_KEY=***
        # Backend image ships its own HEALTHCHECK (uwsgi_curl against /api/health/<token>).
        # Override only the timing so cold starts gate quickly instead of waiting out the
        # image-default 60s interval; the health command + token stay upstream's.
        healthcheck:
            interval: 5s
            timeout: 3s
            retries: 12
            start_period: 10s
        volumes:
        - "{{ fs.app_data }}/data:/data"
        networks:
        - portal