During stress testing we ran 100 concurrent logons and then logoffs with the first 84 user completing this test run in 20 - 30 seconds. During this time both Portal ap1 and ap2 hit 100% utilization on the servers. After the 20 - 30 seconds the portal aps CPU goes idle and the
remaining 16 users out of the 100 hit a 120 second delay before there request returns processing the test and completes. During this 120 sec delay I see the aps CPU go idles until the 120 sec time expires.
What is causing this 120 sec delay?