I've upgraded with success from NNMi 9.11 to NNM9.20 two systems with Application Failover. Platform is RH5.5. Then, in the Incident Browser I can see NNMi Health Critical Event with the following explanation :
"The average garbage collection pause of 53,00 seconds for the 'PS MarkSweep' collector is above the recommended maximum of 10 seconds."
The collection pause is variable and can grow over 90,00 seconds!