I think the idea of monitoring on the deviation from an average runtime is spot on. I have used this methodology extensively to always be one step ahead of the users and bosses, and to stop problems before they become BIG problems. I have used the automated email to call a user running the query and have some fun (sometimes I love my job:exclamationmark
. I have saved countless hours by eliminating problems as they occur or quickly finding the root cause of a problem that occurred in the middle of the night or weekend, which has more than paid for itself multiple times over with just my sanity alone!
Since I am using a supported tool, SQL Sentry's Event Manager, I also have a graphical display showing me the interaction of all the events fired off in SQL or via the Task Scheduler in a cool calendar format, allowing me to drill in and see what was affected and/or caused the problem.