Today I've noticed the presence of an offset on PR_TX (Fig.1). The "ball" generated by the 2-D istogram of PR_TX_INPUT and B4_12/112MHz_mag is skewed at a positive angle, suggesting that the maximum of the error signal's fluctuation should be around 0.006 mA. Both 12 and 112 MHz sidebands show agreement on this, which isn't the case for PR_TY.

Interestingly, this offset was not there a week ago (Fig.2).

I've tried to slowly change the setpoint of PR_TX to recover this offset. This was beneficial for the sidebands power, obviously, but the effect on the sensitivity is negligible (Fig.3) and it's invisible with respect to its normal fluctuation.

The offset has been introduced online, nothing in the automation has changed. Given the negligible effect on the interferometer, I see no harm in leaving it online for the rest of this lock and observe any possible difference over a longer period of time.