Two peak-pairs show up: (#1) 22.80x and 23.20x Hz (where x=0 +/-2), and (#2) 22.70x and 22.30x Hz (again x=0 +/- 2). Figure 2 in particular compares
the noise in B1 in different long locks from Feb 22 to now, it is clear the amplitude of the peaks changes, and it seems they were not present on April 30 and before, but it is not possible to conclude on this point.
Let us focus on pair #1 which is more evident and precise. It "jump to eyes" (salta agli occhi) that the peaks frequency are equally spaced (0.20x Hz) from 23.00Hz (as well the two peaks inpair #2 apears equally spaced from 22.50Hz).
In Figure 3 you see the very low frequency noise of B1: the most intense dark fringe spectral component is a peak at 0.2 Hz. This, according to Paolo, is the first pendulum mode of all the SA filter chains.
To check if the 23.8 and 23.2 Hz pair could be a modulation of this 0.2Hz noise by one hypothetical "carrier" at 23Hz, I played this exercise: I created a 23Hz sinus signal which I amplitude modulated it (i.e. multiplied) with the dark fringe signal, and then I computed the coherence between this signal and the dark fringe. Figure 4 shows that indeed there is some coherence at the two peaks.
A tentative conclusion: the "lines" close to the VELA seems not to be associated to some "environmental noise like cooling fans", but instead seems to be the result of some non linear coupling mechanism of the 0.2Hz noise with one 23Hz carrier line.
The noise model is really too vague. One possibility could be coupling of one of the calibration lines (which shake the mirror) with the 0.2 Hz low frequency mirror motion. Actually it seems we have no calibration line at 23Hz. Might be a combination of different calibration lines ??... we need to investigate deeper.
