A problem existed with VmPoolMonitor throwing a NullPointerException while starting a guest that in turn remains down with its images locked. There was a chance that the prestarted pools' VMs ended up in an "image locked" state if turned off and the engine decided to turn them on immediately. This was caused by a race condition in async task management
The race condition is now fixed using proper locking of the shared resources. This means the prestarted VMs can be freely turned off and they will not end up in "image locked" state.