As reported in #50914, the spikes were present again tonight, then they disappeared during the activity, then they came back without any intervention on the ITF:

• what has happened in the time frame of the disappearance is the following:
  • many optics were aligned/misaligned, but aside a very few urads (or fractions of) for PR and SR and the usual alignment by the drift control for the arm cavities, no macroscopic changes were done; also, the translation of the PR did not change a lot in absolute value (Figure 1, more info below);
  • we suffered three unlocks of the INJ system; the first one in particular shows the biggest change in the working point (for example, the Z position of the MC) (Figure 2, more info below);
  • we relocked the ALS on both arms with the reallocation on the mirrors; nothing was changed on the North Arm, while the BPC on the West Arm was open at the beginning, so it was closed before relocking the arm; the drift control for the beam direction was engaged during the time of the ALS lock; as a test, I disabled manually the BPC on the West Arm while the IR was locked without the spikes, but they did not come back;
  • the spikes did come back eventually, very slowly, during a single, non perturbed lock (still ongoing); they showed up at the same time on both arms, but initially they were a bit more visible on the North Arm, possibly because the SWEB alignment is a bit worse w.r.t. the one of SNEB;
• in Figure 1 there is the overall trend of the phenomenon, alongside the position of all the optics: it can be seen that when the phenomenon worsens, the local controls angular signals do not get particularly worse with the exception of the ones for the PR translation stage (which controls the overall beam direction) and the BS angular controls (which do the same, but only for the West Arm);
• in Figure 2 there is the trend w.r.t. the INJ system: here a couple of things can be noted:
  • the strange behaviour of PSTAB_AMP_CORR, which was stable during the spikes' disappearance, then had a jump around 20:00 UTC, followed by a slow trend; I am no expert of the system, so I don't know the possible meaning of this; a longer trend of the week is in Figure 3, maybe it can be helpful; the week before (Figure 4) seems much more quiet on this matter, but still the correlation is unclear (no spikes on the 16th Feb, for example);
  • the three jumps of Sa_MC_F0_Z are due to the three unlocks, followed by the scan of the RFC resonance; no correlation is evident, and even some mistuning of FmodErr never had this kind of effect on the arms powers, to my memory;
  • the corrections of the BPC are slowly diverging; the DMS pointed me to look at _UV in particular, which went above the 4.5 V threshold (Figure 5), but even much higher values did not cause this effect in the past;
• in Figure 6 there is a snapshot of the arms loop: it can be seen that the accuracy worsens, but the corrections do not change;
• in Figure 7 and 8 there is the trend of the powers of the green beams (transmitted and reflected) and the local controls used by the BPCs; there is some trend on the NEB_MONI signals, but not on the WEB ones.

In conclusion, the major effect was seen on the PR translation stage (and the BS local controls) which could point to something related to the pointing of the main beam, but correlation to INJ signals was not straightforward to me; beside this, it does not look like it is related to the position of some optics, or to the arm control loops; at a first look I could not find a correlation to the green beams either, but a switch-off test could be definitely useful, also in view of later stages of lock acquisition where ALS will not be needed anymore.

Just a remark: the worsening of PR and BS local signals in the trend data likely shows that the noise on the arm powers is affecting the dither signals, built demodulating the power itself. The dither signals are the ones in loop if one look at the very low frequency. The local signals are out of loop, and they register the increased noise on the in-loop signals.

I had a closer look on the spikes on Feb 27 and 22.

Figure 1 shows an example of a few spikes from Feb 27. One can note that they only happend when B4 is large (and conversely B1p is small), that is when the BS moves to a dark fringe position for MICH.

Figure 2 is a zoom. The glitches are visible on B7, B4 and B5 and less clearly on B1p. On B7, B4 and B5 there have the same sign, the power goes up and down at the same time. For B5 the interference is only constructive, the power goes only up compared to the level far from the glitch, for B7 the interference can be constructive or destructive. For B4 it is hard to say as the base level fluctuates much more than the glitches due to the MICH fringes, but they give me the impression of being both constructive and destructive intereferences.

The glitches seem to be also correlated with the regular 1.7Hz oscillation visible on B2. The glitches only happen when the B2 power is high. I don't know what is the light on B2, I expect PR is misaligned in this configuration so there is no real "B2" beam reaching the B2 photodiodes.

Figure 3 is an exemple from Feb 22. Again the glitches happend when the B4 power is high and B1p power is low (MICH dark fringe), and at the same time the B2 power is high.

Figure 4 and 5 show a spectrum of B2 at that time. The oscillation on B2 is very regular and mostly at 1.7Hz. I guess this a dither line. Which mirror is being modulated at 1.7Hz? This dither seems to be putting some mirror in an interference position at each cycle.

One thing to check if this not due to a combined effect of PR and SR misalignment. That is that SR compensates the misalignment of PR, and causes the beam to be realigned in the CITF, but with a lateral shift. This could explain why there is a beam reaching the B2 photodiode. One way to check would be to move PR in TX. If that is indeed the explanation, then PR probably needs a TX component in its parking position in addition to the TY misalingment.