Dave Mason discusses monitors and what happens when they fail:
I was reminded of this recently in my little SQL Server world. I have a number of garden variety alerts set up, plus some other more custom monitoring stuff, which is mostly tied to DDL triggers and event notifications. The one thing all of them have in common is database mail. You can probably guess where I’m going with this. Yep, database mail stopped working. A couple weeks passed before I realized it. Fortunately, out of all the alerts I should have been notified about, none of them were serious.
How would I prevent this happening in the future? I guess I could build another system to monitor my monitoring system. Something like System C, which monitors System B, which monitors System A. But where would that end? System D? System E? Where should the line be drawn? I don’t know that there’s a right answer here, although admittedly, the farther into the alphabet you get, the more absurd it sounds.
At some level, process becomes the answer. In my case, not before I create a few more systems…