so vms wouldn't be cross hypervisor but I could taint and use tolerations the nodes to keep vms from getting scheduled incorrectly. I don't need vms to move between the x86 nodes and arm nodes.

Also AMD and Intell need to be separate clusters.

never heard that one before. I think it's fine if vms don't transfer between arm and x86 but I wonder if the control plane would support it otherwise or if the harvester control plane would act unpredicatbly