Koji had the good idea of trying to measure the motion of the POP beam, and feeding that signal to PRM yaw to stabilize the motion. To facilitate this, I have installed a 50% beam splitter before the POP 110/22 PD (so also before the camera).
Before touching anything, I locked the PRM-ITMY half-cavity so that I had a constant beam at POP. I measured the POP DC OUT to be 58.16 counts. I then installed a 1" 50% BS, making sure (using the 'move card in front of optic while watching camera' technique) that I was not close to clipping on the new BS. I then remeasured POP DC OUT, and found it to be 30.63. I closed the PSL shutter to get the dark value, which was -0.30 . This means that I now have a factor of 0.53 less light on the POP110/22 PD. To compensate for this, I changed the values of the power normalization matrix from 0.01 (MICH) to 0.0189, and 100 (PRCL) to 189.
After doing this, I restored the ITMX and am able to get several tens of seconds of PRMI lock (using AS55Q and REFL33I).
I found several QPDs in the PD cabinet down the Y arm, but no readout electronics. The QPD I found is D990272. I don't really want to spend any significant amount of time hacking something for this together, if Valera can provide a QPD with BNC outputs. For now, I have not installed any DC PD or razor blade (which can be a temporary proxy for a QPD, enough to get us yaw information).