You tried the normal things like rebooting the node?

My best guess is that there's some socket file/lock that halting it from progressing and the reinstall fixes it?

But you're trying to intentionally force a reboot/shutdown and figure out how to bring it back up?

so you have a bunch of empty or corrupt state files, as well as corrupt binaries

So I tried also microk8s and it I was not able to break doing the same. They use dqlite instead of SQLite. So could db be corrupted ?

If you shutdown a node ungracefully, (pulling the power, or even just shutdown from the cli while it isn't fully established yet) the OS/filesystem isn't equipped to handle that kind of distruption.

Unless you're using something that's stateless/container based for your OS like FedoraCore OS or Suse Transactional server, elemental, etc..

The ungraceful bit is means more than just hard hard reboot. Hard reboots should still be graceful. It's possible to recover from a loss of power event, but not guaranteed with just normal Ubuntu.