Detail:
Since PRMI 3f lock is not stable, I tried to lock PRMI with carrier on rsonance with 1F signals and checked the 3F signal.
To lock PRMI with a cariier on resonance, I just flipped the sign of PRCL control and can lock without any problem.
At that time, beam power on POP PDA1 was over 2mW (if the calibration factor in FB is correct), while AS power was about 1.1 mW (fig1).

First, I compared 1f signal and 3f signals when passing the resonance (fig2).
Both 1f and 3f signals don't have strange error-signal shape as reported in klog15181.
I also checked if the zig-zag shape can be reproduced when I rotate the phase and it can be.
So, phase seems fine.

After that, I checked the 3f signal when locking PRMI with 1f signals (fig3).
Power on POP was fluctuated maybe due to the oscillation between S-pol and P-pol reported in klog9325, but it doesn't matter at this moment.
REFL51Q signals are almost zero but REFL135I signals have an offset.

Also, I checked the 1f signal when locking PRMI with 3f signals (fig4).
In this case, POP17 signals (MICH 1f signal) have a huge offset: that means MICH is not locked in the AS dark condition.

Currently, I'm not sure the reason of this offset but one possiblity is RFAM, so we need to check it.