Following my talk at the Systems meeting, I've added some pdfs to the DCC entry with 'publicity plots' of the improvements due to the BRS:
G1601529: Tilt- and Sensor-correction filter tuning at LHO.

The figure of merit I apply to these plots is that we want to reduce RMS velocity. I believe this is the correct figure of merit 
from DC up until at least the lowest Quad resonances. This saves us the worry of trying to compare bumps and dips in spectra. In all the 
following plots, the dashed curves represent the high-to-low RMS of the solid curves of the same colour. Most of them start from some 
cut-off frequency, to isolate the low frequency contributions. In general, the output of all inertial sensors can be regarded as noise 
below 0.1Hz, since it's either uncorrelated or it's something that will be common to the ends and the corner.

The only new figure here 'Sensor correction vs not' suggests that the noise injection from sensor correction, about 8e-8m/s from 0.1Hz to DC, 
is lower than the potential benefit of sensor correction, about 9e-8m/s from 0.3Hz to 0.1Hz, even with high winds (25mph) and very low 
microseismic motion. With higher microseism and lower wind, this ratio will only improve. Still, during very quiet times, we may be better 
off disabling low-f sensor correction everywhere.