have you got a chance to review the complete PRPC logs to see what operations are taking time?
1. Any long running queries?
2. Whats the requestor count? Can i get the environment details? Have the capacity planning is done for the environment, like how much the system is capable to respond for the number of requestor / other operations?
3. Are there any frequent RD's running which are taking huge time?
4. Depending on the kind of operations, have you reviewed the alert logs, to see what is frequently hitting the system?
5. Why is thread dump being triggered?
Whether any operations like activity / any thread which is stuck and executing for long time, making other many threads getting piled up?
6. DB Intensive operations?
7. How many users are accessing the application? Requestor Count?
8. How many nodes? Environment details?
It goes on..
I would suggest you to send us the Pega-Rules and Alert Logs, from startup till application going down.