I asked Valerio Boschi what was the status of the WEB motor crates: he confirmed that they had been turned OFF via the "smart power strip" (no power applied) thus leaving the switches on the front panels in the ON position.
We know that this condition breaks only one of the two mains wires, whereas acting on the front switch interrupts both, avoiding the coupling between the noise and the WE mirror.
So this morning we went to WEB and first turned all three affected crates switches to the OFF position (08:04:00 UTC approx.), noting a noticeable reduction of noise in Hrec in the 40 - 60 Hz band, as expected.
Then we selectively put back to the ON position the switch of one crate at a time: the first at the top at about 08:08:00 UTC, noticing no difference (the noise remained low), and then the second in the middle at about 08:12:30 UTC.
In this case we seem to have seen a slight increase in noise, but the interferometer was noisier in general, so it's hard to assess that, partly because right after that the interferometer unlocked at 08:14:30.
Eventually, we left all three front switches in the OFF position (thus minimal noise in Hrec), waiting for a new lock.
In fact, we would also like to try under this condition to physically disconnect the rear mains power cables, thus also breaking the safety-earth connection.
We will then decide what is the best configuration to leave.