Since PRFPMI_RF_LOCKED seems stable, I tried to lock PRFPMI_RF with 15W.
However, lockloss happened many times during increasing laser power, which can be passed in the previous time, so I gave up to increase the laser power today.
One possibility is glitches when changing IMC CMS IN1 gain during increasing laser power but not so sure.
Further investigation is necessary.

Note:

Since we replaced CMS (klog35394), it might be due to the indevisual difference of CMS.
So, we need careful check of the difference of the signals from the previous trial.