I checked the Service Request.
There are a lot of "moving parts" for investigation. I see discussions on OAM SSO, Load Runner tests, etc.
There appear to be some outstanding action plans that should be completed.

1. Update to the latest patched client tools
2. Clean diagnostic dump after the issue reproduces to assist in narrowing the issue/ check for any crashes, etc.
3. Questions about agent schedule/backlog
4. RPD configuration for Exadata Oracle DB source needs some updates.
5. Checking for initialization block failures, hangs

6. Checking ulimits.

Previously, a 'broken pipe' was reported, which I see in the logs attached here.. that can indicate the nqsserver crashed (as Moustafa) mentioned, but there is no .out file attached here to correlate that.

Next time you restart, notate the PID of the obipsX, obisX, when the issue occurs, check if the PID changes (that is a easy clear indication of crash), in addition there should be crash reports generated in the component logs directory.

Once the issue is narrowed, the thread can be updated; otherwise, it may be guess after guess here without proper data to perform root cause analysis.

The service request is also 24*7, which can be helpful in some cases, but also not efficient if there is no traction on the issue, or new lines of thinking without follow through. I would suggest you align it with someone in your timezone. The assigned person has a team they can collaborate with, but this really needs a thorough methodical review.

BTW, you should change your netstat command should be filtered by the nqsserver port (default: :9514, but you can check with status.sh -v command)

There is an SR open, all the above has been suggested already.
There is progress being made, all will be summarized post solution.