I went back to the end station to look into why the frequency comparator readback has not been reliable.  There was 1MHz oscillation on the readout of the PD that is similar in size to our beatnote, which was confusing the frequency comparator.  The oscillation goes away when the noise eater is on, and the frequency readback works well.  For now I have left it on, although this isn't a long term solution.

In alog 6669 I came to the opposite conclusion, that the nosie eater had to be off to avoid oscillations at 1 MHz that confuse the frequency comparator. A high pass in the path to the frequency comparator might take care of the problem.