When upgrading harvester instead of using the standard upgrader, is it possible / supported to manually wipe each node individually and reinstall using the newer ISO? You would need to give nodes new names and remove the old ones from the cluster but this would give more control until https://github.com/harvester/harvester/issues/6145 is implemented.

I just removed validator and started upgrade (as I had no choice as our firewall is running inside cluster...)

Does it just get stuck nodes until you move single replica volumes out manually?

nope. works automatically until doesn't (but that is another story)

Did you have single replica storage that the upgrade checker failed on? How did harvester handle the nodes with VMs it couldn't migrate during the upgrade?

well. I had already shutdown most (because of that start problem). In my understanding it works how you have defined in vm config Maintenance Strategy. Only problem is that this config is useless (as validator does not let you start upgrade).