Re: A real puzzle - Electromechanical fault
 

Small spikes aren't a problem, I see a lot of brown outs that don't affect the radio or cRIO forcing them to reboot.
One toe below 4.5v and the protected outputs are gone though, and I've seen that happen from sharp spikes.

Monitoring the battery voltage ourselves in code has worked out when we have had a big draw device, such as a flywheel.
The FPGA code also monitors the battery voltage and reacts when it drops below a threshold voltage, disabling the PWM/Relay outputs.

P.S. I made a correction to my earlier statement, because I'd said that the PDP cuts the regular 12v outputs. It doesn't, but the effect is the same when the PWMs get disabled.

Attached is the circuit diagram.

Re: A real puzzle - Electromechanical fault
 

While I'm sure you've checked this, it could always be a problem: could you be running the two CIMs in each CIMple box against eachother?

Re: A real puzzle - Electromechanical fault
 

They are running 4 cims on 4 cimpleboxes... So I don't think that's the case.