In the top panel of the attached, red is the ETMX OL PIT, pink is the dark noise (this is ETMX OL PIT when there's no light on the OL, but a fake offset of 9900 counts was applied to OL SUM so the scaling of red and pink become the same).

The RMS of the red line for f>0.1Hz is about 0.15urad (that's about 0.5 urad pk-pk). The dominant hump at 0.4-0.5Hz is about 0.12urad RMS, but it seems like the microseism is adding another 0.9urad RMS, resulting in sqrt(0.09^2+0.12^2)=0.15urad RMS.

Dark noise is well below the PIT signal up to a few Hz.

Green is the OL SUM normalized by OL SUM DC, and then scaled by H1:SUS-ETMX_L3_OPLEV_PIT_GAIN. If  the green stays on top of the red for some frequency band you'd be suspicious that the OL quadrants are measuring some noise that is incoherent across the quadrants (shot, dark), but that's not the case.

So OL seems to be measuring the angle at least for f<1Hz, probably we don't have to worry too much about injecting more noise than we want by using OL damping. Even if the actual PIT noise floor is as big as green (which is not likely), the RMS due to noise injection is not bigger than brown dashed curve.

OTOH, at least for the time when the measurement was done, it seems like the damping for 0.4-0.5Hz hump is not good enough, microseismic should be attacked too.