I did some calculations on the displacement-noise properties of PR3 by using the yaw-noise data measured in April this year from Akutsu-san on the respective OpLev. These calculations should show how the outcome of a length-sensing OpLev-QPD would be influenced by the noise of itself and its surroundings.
It seems by using a lens with f=500 mm we would reach the requirement of measuring at least 1um shift of the mirror. The big feature around 1Hz is due to the Kumamoto earthquake that happend in the period of data-taking.

However, this are just short-term spectra and still it is not quite clear whether such a lens can be installed.
please note that by using a lens with a smaller focal length, the noise spectrum would increase with a factor (900-f)/f, where 900 is the distance between mirror and lens (in mm as f is given in mm too).