I've done a couple of things since the earlier post, none of which have reduced the noise level.  First I put some lead bricks (and a couple of SR560s once I ran out of lead bricks) on top of the cover.  I put a sheet of foam between these and the lid, in the vain hope that this might damp some accoustic coupling through the top.

Next I put in two beam dumps behind the flat mirrors in the cavity.  These should really have been in already since there is a small amount of transmission through them.  I can't get to the other two transmitted beams because of the mirror mounts.  We might need mounts that allow access to all beams in and out in order to block these.

I went back to looking at the locking signals again.  The slow feedback on the CCW loop appeared to be oscillating a bit and I turned the gain down till it stopped - a factor of 10 below where I started. I also added the boost on this side.

I then cranked up the gain on the CW loop as well, and added the boost.  I turned it up till it started oscillating and then came back down by a small amount.  Still no different.

I find it intriging that the noise is much lower with the FM modulation turned off to the AOM.  It seems to me that this might point to it coming either from the locking loop for the AOM, or a mechanical noise source, since it doesn't appear to be the Tektronix at this level.