When a bridge interface is configured and used for Apps, on a reboot Kubernetes initializes before the network does. It does not see the bridge as existing, and fails. Through the GUI it is then possible to change something, reinitialize Apps, then change the setting back and it will start up fine with the bridge.
This issue has been outstanding since Cobia and was reportedly fixed, but many users have found the issue has persisted, even after following the recommended sequence of deleting the bridge, rebooting, and creating it again (myself included). The issue persists in Dragonfish 24.04.0, and as far as I can tell it’s not even possible to follow that sequence anymore. You can’t delete the bridge because Kubernetes is “using” it (even if the pool is unset and the service is not running).
There have been numerous forum posts and bug reports about this in the old server. I’m starting a new one here in case anyone has found a solution to it. This is the most recent JIRA issue I’m aware of (Log in with Atlassian account).