If a VM is suspended and the save file becomes corrupt somehow, either because it was never saved properly or otherwise, Nova cannot recover. It will continue to limit the user to "resume" or "terminate" with no option to just reboot without trying to load the save file. The VM will be in this state forever until an admin manually intervenes. We can do better, as the condition is completely recoverable.
The fix is to add proper exception handling to the driver.resume() call such that it catches this case, so that it can clear the saved file (equivalent of 'virsh managedsave-remove') and simply start the VM normally. The user should probably be notified when this occurs, to the effect of the resume having failed but the VM being rebooted anyway.
No activity for 575 days, so changing state from In Progress to New.