Unstable B Axis

I have a 5 axis MVJR Mori Seiki that is retrofitted with LINUX CNC.

Most things are working except the B rotary axis has an issue. It moves as expected.
But, if the machine is on, just sitting there the B axis will occasionally jump a short distance.

If I turn the servos off for a long time (several days) but leave LINUX CNC on, and then Turn the servos on, the B axis will violently jump a long way about 20 degrees.

My guess is that something is wrong with the encoder or wiring that is falsely signalling that the axis has moved.

Does anyone have any ideas for other things that could cause this?
Or how to diagnose the source?

I am thinking of maybe getting an analog continuity tester and testing the connections to the rotary axis encoder.

Use halscope or show hal configuration to monitor the encoder and the PID settings for that axis, make sure it is not caused by PID wind up.

Thank you for the advice.

site:https://forum.linuxcnc.org/ PID Windup

This search returned some interesting threads, I will work with that tonight.

That sounds more hopeful than tearing everything apart looking for a bad connection.

I read the articles I could find about wind-up but they were mostly about spindle.
I am still searching but don't know what to look for exactly.

Can you tell me what clues I should look for and what PID setting I should expect to be involved in the solution?