during a service group failover/migrate, if the application can't be shutdown (within the resource offline time limit) shutdown gracefully, agent CLEAN script will be called to forcefully (like kill) offl;ine the resource. There are multiple agent/resource offline timeout/limit attributes which can be tuned. If even clean script os not able to "kill" the process(application), the node will be rebooted to prevent data corruption.
There are many agent/resource parameters, Run the commands below to view them
1. haagent -display <agent name>
2. hares -display <resource name>
the details of the above mentioned parameters are in the VCS admin and agent guides.