We had a lockloss at 50W due to what looked like the ISS 3rd loop, so I re-measured with higher recycling gain.  It seems that when we recover recycling gain by moving the PRM, we get some loop interactions that reduce the gain of the 3rd loop right where we need it.  So far, just increasing the gain of the loop has been enough to keep things under control.

Attached is a set of measurements of the 3rd loop, with the state of the IFO noted in the magnitude legend on the upper left.  "g" refers to the gain H1:PSL-ISS_TR_GAIN. 

Once, while the 3rd loop gain was -1.5, it started to oscillate.  I put the gain to -2, then down to -1.  I think it liked -1 better at that time, and that's how I left it for the rest of that lock, but -2 could also have been okay perhaps.  I'm not totally sure.  Perhaps we need to give this thing a UGF servo?