Lindsay Hill comments on his blog:
Network monitoring is often a reactive process. Engineers see an alarm, and take action. Alarms mean someone broke something, and it needs attention. Deal with the alarms, and all will be well. No alarms == Good. But every now and then no alarms means that monitoring is broken. Here’s an example of where this happened, and what I do about it.
Unknown unknowns are a bad thing in networking monitoring. Never assume that a quiet period means no problems. Always verify that things are working correctly. Read on for Lindsay’s examples.
Read more at: No Alarms May Not Mean No Problems