Occasionally I have a single VM go into an "?" unknown state and the only method of recovery is to kill the PID on the host.  This happens once or twice a week to 1 out of 500+ VM's.  There is no pattern to this and VM operating system doesn't seem to make any difference.