We've looked at the times during shaker injections. We suspect that 101 Hz is not caused by the RFC minilink or by some structure on the suspended bench, but perhaps by some structure attached to the chamber wall inside (is there some lip where the minilink meets the wall that could have its own resonance?).

• 101 Hz is not a resonance of the RFC minilink. We see no resonant peak in the minilink accelerometer. While there is coherence around 101 Hz between RFC Err and accelerometer, there is less coherence right at the 101 Hz peak seen in RFC Err. 
• 101 Hz is not a resonance of some structure on the bench. Noise in Sc_MC_RFC_Err_Post scales as 1/2 the RFC minilink motion (IB_BD_ACC) and the same as chamber wall motion (IB_ACC_X).  Unless there is a mechanical short between the chamber wall and the suspended bench, motion of some bench structure (optics mount, etc.) should be attentuated relative to chamber wall motion.
• Additionally, this doesn't look like two structures moving/modulating scattered light; the motion of the chamber wall/minilink is large (10^-4 m/s²) compared to wavelength, so we would expect broad noise peak, not the narrow line we see in RFC Err/DARM.

I attach spectrum during no injections and two different shaker injections, showing the accelerometers and error signals mentioned in original post. In the plots: purple is no injection, green is 75-125 Hz shaker injection, and blue is 95-105 Hz shaker injection.

I noticed that ~60 Hz peaks were also excited in RFC Err during these IB tower shaking tests even when we were only driving close to 100 Hz (blue line). As Bas has noted, these peaks are some of the worst for DARM.

In that confirguration, I think you were locked in recombined but not CARM to MC.

In the morning, we mostly worked in the states CARM to MC and LOCKED SSFS.

It would have been interesting to do it also in these 2 ITF states.

By the way, looking at the data it seems we can find coherence with the RFC error signal (mostly with accelerometers located north side of IB tower).

Even the SPRB one is coherent with the RFC error signal in several regions.

Two additional observations:

when shaking the IB tower tank several peaks excite in the RFC signal (Figure 1), the 62Hz was already noticed by Terra.  Among them the only one that is coherent with IB tower tank accelerometer also in quiet condition is the 101.Hz. 

the shape of these 101Hz (Figure 2) and 62Hz (Figure 3) peaks match very well the shape of the peaks that got excited when Maddalena injected noise on IB angular controls ( 41956): see Figures 2 and 3. These peaks should be mechanical modes of SIB1 last stage suspension (quoting Paolo)  (broadband noise injected on SIB1 coils should definitely prove it). Figure 2 also shows that also the shape of 101Hz peak in Hrec maches. Figure 3 shows that instead the structure in Hrec around 62Hz was something else.