systemd service file should include a "Restart=always"

When the client push is configured but the service is not running, this causes permanent browser issues as the browser always tries to establish a wss connection to the push service. You can see that in the browser console.

As a side effect this also delays the connection to the signaling server in Talk when entering a chat room ... up to messages that the signaling connection does not respond in time.

I found that, when the complete server VM is started, the notify_push service can start before the database is up. That leads to an error and notify_push often does not run.

The additional command

Restart=always

in the [Service] section should restart the service in that case until the database is up and the service is running.

Maybe that should be added in the documentation (sample service file).

Locate the notify_push systemd service file and inspect its [Service] section. Verify the service behavior when the database is unavailable during startup, then confirm that the unit retries until the service runs successfully.