Koji is working on PRMI locking, and while he was doing that I glanced at the oplevs' spectra for the ITMs and PRM.

I found that when the PRMI was locked (for only 1 second or so max lock time) on the 55MHz sideband, and the error signals show a big peak around 400Hz (definitely audible in the control room), the only OpLev that I see a similar peak in is ITMX pitch. 

In the plot below, I have grabbed a time when the PRMI was flashing as the black reference traces, and then a time when the PRMI was locked as the active traces.  You can see that there is a similar peak in both REFL55I and ITMX_OL_PIT when the cavity is locked.