Is it possible to create "Host lost" discovery triggers without scanning an IP range?

Currently we have hundreds of machines that come up and go down throughout the day. When the machines go down, they sometimes interfere with aggregate checks on their group -- their stale data...