First I have tried bruco, and adding some risky but useful control channels, but that failed with an out of memory issue half way through.
However looking at some figures that bruco did manage to produce SIB2_B2_8MHz_I looked intersting.
On figure 1 is the coherence and spectra with PRCL and SSFS. The coherence is at the ~30% level between 1kHz and 3kHz, in particular there is coherence at ~2kHz where none of the spectra shows any structure. So this could be dark or quantum noise coupling from 1 DOF to the other. It would be interesting to do a full set of noise injection into PRCL, SSFS, DARM and MICH in that frequency region (1-3kHz), and see which direction the noise propagation is (from DARM to SSFS and PRCL, or the other way around).
Given that the SSFS is the only loop that has gain at 1kHz, it is most likely origin of the noise. This is also supported by the noise budget (figure 2 and 3) that gives a contribution of B4 56MHz quantum noise in that region. But I wouldn't trust the budget too much, as I had to change the coupling of SSFS noise by a factor 30 to match the model to the measurement.
Having a usual set of noise injection into each length DOF to measure the loops would be also useful. As currently in the noise budget I have major doubts about the PRCL and SSFS loop and noise propagation in the model.
