Hopefully this helps someone else in future.
We solved this by restarting the hosted engine. That got us partial control of the host
via the engine although the host refused to go into maintenance until VDSM was restarted
on the node. Once in maintenance it was rebooted and all back healthy