As suggested by Raffaele, I made a scan of the ML_TH_CORR with the IMC unlocked.

I put the rampeauto of the IMC in "manual",  misaligned by 30urad MC_TY, then:

1st plot: made a slow scan of ML_TH_CORR, both increasing and decreasing. While scanning, the PMC cavity lost lock many times. We can see the Slave laser PZT corr following the ML frequency variation quite smoothly, while this does not seem to be the case for the PMC piezo correction.

2nd plot: made a slow scan with PZT_CORR and EOM_CORR cables unplugged from IMC rampeauto. Did not see striking difference.

We zoomed in on one of these PMC unlocks: 3rd plot, where we can see a fast glitch occurring on the power delivered by the SL in coincidence with the unlock. Zooming further in (4th plot) the glitch seems to be due to some optical cross-talk with the PMC reflection (TBC).

No conclusions at the moment.

The IMC was relocked at the end of the measurement.

EDIT: the ISYS configuration did not change wrt the past days, i.e. we have the new fibered amplifier replacing the SL-Neovan,but the SL is still injection locked to the ML, and its beam is blocked before entering the Neovan. Also, before making a large scan with ML_TH_CORR (reaching out values as we saw in the past days), we made a scan comparable with that made on friday evening (40mV). This is reported in the plot 5, where we can see that the PMC loses its lock while decreasing the correction, but not when putting it back to the original value. Plot 6 is a zoom of the lock loss of the PMC.

In plot 5 we can see indeed a glitch in 50Hz-sampled error signal of the SL in coincidence with the lock loss, but not in the related fast channel (plot 6), perhaps because of the high frequency noise.