Only a thought. Did you configure memory or CPU reservations on any of your VMs? In this case you may have an increased slot size which limits the number of available slots and therefore the number of VMs which may be powered on in the cluster. You can verify this in the HA's advanced runtime info.
André
