After the 76623 frequency noise injections, we left CARM control just on REFL B (H1:LSC-REFL_SERVO_IN1GAIN = 12dB) rather than the nominal  6dB on both REFL A and B. Kept the REFL B only config from 17:39UTC to 18:58UTC.

Plan to look for coherence between DARM and REFL A to check for frequency noise on REFL B (when REFL A is out of loop, acting only as a witness). Plot attached shows no coherence in 100Hz to 1kHz region above 0.04.

Test was maybe redundant after Francisco's 76533.