It was a one time recorded incident. It may be more frequent, but you do not discover such things easily when your workstation is not noisy.
I just wanted to let you know that it is possible. The culprit was the HFSS solver left in RAM after the session was finished.
It was continuing for the day and night, if I am not mistaken, or longer. I was surprised, actually, by high load when I started another session, and was just editing, without any simulation.
I.e. the 4 processes on the screenshot did not require a license! But I guess, they may live a trace in your spy-log as non-licensed processes, since I was starting an 8-thread simulation around that time.
Killing the process fixed the things.

BTW - speaking of licensing, your licensing system is awful, and unreliable. Licenses going stuck on the server forever with any minor distortion of IP-connection. It forces people finding ways around. Today your legal dept intimidate people for doing that, but one day a court precedent may reduce all your future attempts of calling for license terms obedience, since the licensing is simply un-re-li-ab-le.
Just make the normal licensing mechanism. Normal, I mean, which withstands RJ45 disconnects, and VPN connection loss. Secure session tokens, and machine id-keys were invented decades ago, but you still rely on TCP ESTABLISHED state, and if that is lost, all the work for the day is ruined.
- yes, I had to go and restart the server again a day before. but I can't do that on remote.