I took a long noise injection measurement of the MICH OLG, screenshot and template attached here.
Anamaria had suggested that the ~17.75 Hz lump in our DARM spectrum could be due to BS bounce mode. 49643 has an old measurement of this frequency from before the bounce roll mode dampers were installed.
The second screenshot shows the MICH and PRCL error signals bandpassed with an elliptic bandpass from 15-20 Hz at a time when PRMI first catches lock. The dominant frequency in the MICH error signal is in the right frequency region, 17.65Hz measured by the three cycles shown in the screenshot.
So perhaps this noise in DARM can be fixed with a notch in MICH.