I have made a lot of changes to fix this issue, but think I finally [serendipitously] figured out how to resolve this yesterday.

I actually did two things yesterday. 1. I capped SQL Server memory so that 2GB were free for the OS. 2. I turned off all data collectors and added them back one at a time. There is one more important note: I was monitoring the server that the Utility was running on in addition to six other servers. I have not turned data collection back on for that instance.

I think it was change 2. that actually made the difference, but include both for your information. Since that time, I have only had one "The thread "ExecMasterPackage" has timed out" error on the busiest server. I was getting them ALL the time on several servers.

I suspect the problem is related to the these failures causing the log on the utility server to get backed up and rendering it unable to carry the backlog. Our utility server is lightly horse-powered, since this is a relatively new technology. I think "the Utility" [which is a totally dumb name] requires quite a bit of IO throughput in order to avoid problems. I suspect I increased the conflict by monitoring the local system. I also suspect that the problem is related to DCEXEC executions and stops.

Good luck. Hope this helps!