For example, in situation where host:

* laggy

* network laggy

or another manager <-> host connectivity issues - ovirt-engine can't determine what with host actually:

* host stuck

* host unreachable

* me (engine) unreachable

* other lossy, lags, etc

another chances for enginte to determine host is powered up or not is IPMI - engine tells to another (reachable) host to execute IPMI call for target host (problem host):

here some engine logic, when host is actually powered up - we can wait when management network return, or can reboot (via IPMI) host, mark host as failed, evacuate VM's, etc...

k