CARM gain was changed by rotating HWP when LSC_LOCK guardian was in OBSERVATION state.
Figure in the following link shows the spectra when HWP angles were 188 degs (low REFL power: gray) and 172 degs (high REFL power: red).
https://dac.icrr.u-tokyo.ac.jp/KAGRA/DAWG/Detchar/klog/klog24169

The noise floor at high frequencies changes by one order of magnitude while the peak structure below 1kHz didn't change so much.
So, the peaks, which has a coherence with CARM error signals, don't seem direct coupling from CARM.