Because there is no TimeoutCount metric, and unlike errors, there is no error message either (you get a line Task timed out after x seconds instead).

So when your alert goes off because the error count spiked, what do you do? You look for “error” in the logs, and you won’t find anything! Of course, if you know to look for Task timed out after x seconds then you’ll know, but unless you get time-outs all the time you just aren’t primed to think of that. And since time outs are treated as an error in the metrics, they should have a corresponding error log message too.