Maybe a feature enhancement request, but -
How about a threshold on connection state alarms (And similar alarms) saying the system must be unavailable for X seconds/minutes before alarming? We monitor several ESX boxes, some over WANs, and they occasionally disappear for a second or two, only to reappear immediately. But we still get alarmed (Then an all clear.) Would be nice to say, if the machine is missing for 10 seconds, then alarm me.