We put the preamp output directly into a multimeter and observed the same fluctuating behavior as the DAC channel was changed.

We're bypassing the relay to see if that makes any difference. The old relay wiring (to be bypassed) is shown in the attached diagram. That didn't do anything.

We're looking at filtering the DC output by 5kHz to see if there are any resonances at higher frequencies that might go away. Changing SR560 output for AC path to DC and setting gain to 1 on that unit. Also changing gain in FM31 filter bank from 1E-3 to 1. The results are shown in the attached time series. The channels FM30 and FM31 see the same thing. The only difference is that FM31 goes through an SR560 with a 0.03Hz pole (6dB).

Success by bypassing the DAC bias voltage. We switched to a 300mV bias voltage from a function generator. Doing that removed the causal PD voltage drift induced by changing the laser diode current set voltage (see the last time series). So the issue is some weird coupling into the DAC bias voltage.