Steps:
1. Configure a PA to connect to a secondary ISN (confirm it is properly seen in TSPS and TSIM UI and data collection works). A PA Connect event to secondary ISN will be also seen in the TS consoles. The origin value in source section will show the secondary ISN as expected
2. Change the patrolagent config file for the PA to connect to primary ISN in an ISN cluster the next time the PA is restarted or the machine is rebooted
3. Reboot the PA machine which will cause the PA to restart too. This reboot should cause a PA disconnect event of the agent that should come from the secondary ISN where the PA was connected, and a PA Connect event from the primary ISN where the PA will connect after reboot

 ACTUAL RESULTS:

a) If you see the PA Disconnect event arriving first and then the PA Connect event, then the source origin in the PA Disconnect event will be correct (it will show the secondary ISN where the PA was connected before the reboot)
b) If you see the PA Connect event arriving first to the TS console, then few seconds later you will also see the PA Disconnect event in the TS console, however when this scenario happens the origin value of that Disconnect event is INCORRECT, it will show the server name of the primary ISN when it should show the server name of the secondary ISN. Note. The integrationService name shown in the PA errs log file is correct all the time. Also checked the ISN logs and found that PA got DELETED from the secondary ISN, and it was also properly ADDED to the Primary ISN which is also correct.
When scenario b) happens, we have noticed that this is causing an additional consequence. In the TSPS UI the PA that reconnected to the Primary ISN after reboot is still seen in red even after several minutes despite it is connected to that ISN. If you login to TSIM UI and check the PA you also see it in Red as disconnected and all monitor instances are also Red and MFD, however if you check graph from any instance you see the data collection is really happening.

Reduce the value of property pronet.apps.agent.patrol.heartbeat.interval=30000 in the pw\custom\conf\pronet.conf file of the ISNs (from current 60000 milliseconds to 30000 milliseconds). A restart of ISNs is required for change to take place.

Explanation:
There is an ISN Reconnect/failover time set in Patrol Agent side also to 1 min, hence in case PA detects disconnection quicker than the ISN, it will failover to secondary ISN in 1 min.
Having the heartbeat to 60 seconds also at the ISN side keeps ISN to detect disconnect after 60 seconds.

Both the conditions will lead to a race condition in case of ISN cluster and failover scenarios, so keeping the heartbeat to 30 seconds at the ISN will solve the current situation where sometimes the connect event to ISN was received before the disconnect event from ISN. With this change, basically we are quickening the time when ISN sends the proper disconnect message to AC, instead of waiting for 1 minute.
This should not cause any performance issues. It's just that, instead of waiting for 60 seconds, ISN will now send the disconnect message within 30 seconds.