The attachment shows several plots of the CARM error point (as measured by the monitor on the common-mode board) along with the monitor of REFL9I. Both have been referred to volts out of the REFL9 demod board. The left-hand plots show the situation during O1, when there was –13 dB of gain between the summing-node input and the CARM error point monitor. The right-hand plots show the situation as of last night, when the gain was instead +3 dB.

My initial guess is that the improvement in the error point noise is simply due to the gain redistribution between O1 and now, though I have not gone through and budgeted the noise.