Firstly - the poll interval on all the machines is set to 15 minutes (default) but yet I seem to get alerts at very random (sometimes less than 15 minute) intervals.

Secondly - I get notifications of Windows machines being close to 100% CPU utilized, but when I actually look on the box I see numbers way less than 50%.

Thirdly - when I check the values retrieved during each poll (which I get my clicking on the CPU Gauge for this specific machine) I see quite reasonable non threshold violating numbers for each of the 15 minute intervals.

I am attaching some screenshots to demonstrate my point.

Take note all the Windows boxes are Win 2K8 R2 machines.  And this only appears to occur on some of them at random intervals.  Also all machines are cloned from the same base VMWare template - so no differences in Firewall policy etc on them.