• Thanks for showing this to folks and including your findings!

    A couple of quick notes - the Cost Threshold thing isn't really seconds. The BOL topic says it's seconds for a specific hardware config, but that means a long-dead machine at Microsoft. It's the query cost like you see in an execution plan - it's a combination of CPU and IO loads.

    About the MSDB backup history - the problem with this (and I need to phrase it better on the findings page) is that if you use monitoring software that checks to make sure all databases have been backed up in the last X days, then that MSDB query ends up being one of the biggest resource hogs on the system. If you go through the plan cache and sort by total logical reads descending, the msdb backup check usually shows up in the top 10-20. MSDB just isn't indexed well to support most monitoring queries.

    Thanks again, sir!