After talking with Steve, I had a look at the PEM's AA board, to see what the problem was.
Steve said the symptom he had noticed was that the Kepco power supplies which supply the +\- 5 V to the AA board were railing at their current limits as soon as he plugged the board in. Also, he smelled smoke.
I started with the power supplies, and saw that the 2 individual supplies each had a dV=5V, and that the one labeled +5V had the red wire on the + output of the power supply, and the black wire on the - output. The supply labeled -5V had the orange wire on the -output of the power suppy, and the black wire on the + output. Normally, you would expect that the 2 black wires are also connected together, and perhaps also to ground. But at least together, so that they share a common voltage, and you get +\- 5V. However these 2 power supplies are not connected together at all.
This implies that the connection must be made on the AA boards, which I found to be true. It seems a little weird to me to have that common ground set at the board, and not at the power supplies, but whatever. That's how it is.
The problem I found is this: The keyed connectors were made backward, so that if you put them in "correctly" according to the key, you end up shorting the +5V to the -5V, and the 2 black wires are not connected together. You have to put the keyed connectors in *backwards* in order to get the correct wires to the correct pins on the board. See the attached pdf figure.
Since these are internal board connections, and they should not ever be changed now that Steve has put in the adapter thing for the SCSI cable, I'm just leaving them as-is. Steve is going to write in huge letters in sharpie on the board how they're meant to be connected, although since this problem wasn't caught for many many years, maybe it won't ever be an issue again. Also, we're going to move over to the new Cymac system soon-ish. However, whomever made the power cable connector from the box to the board for this AA board was lazy and dumb.
After putting the connectors on the way they needed to be, Steve and I powered up the board, hooked up the SCSI cable in the back, and put a constant voltage (~1.3VDC battery) across various different channels, and confirmed that we could see this voltage offset in Dataviewer. (Kiwamu is hoarding both of our SRS function generators, so we couldn't put in a low freq sine wave like I normally would). Everything looked okie dokie, so I'll check the regular PEM channels tomorrow.
Steve will re-install the board in the rack in the morning.