I tried to recover the IFO lock because the seismic motion calmed down and BS Yaw seems to be shaking less than 0.1 urad. (and I had some time during the observation shift.)

Before the IFO recovery, this morinig we found that PR3 sometimes moved largly, especially in Yaw direction in the LOCK_ACQUISITION state without any touching. Fig.1 shows the time-series of error/feedback singals of PR3 local controls in the BF, IM, and TM stages. As you can see, as for in Yaw direction, error/feedback signals in all stages got large. This may indicate BF Y control kicked PR3. 

Moreover, this kick seems to happened from last night (fig.2). Also, this seems to happen sometimes and sometimes not.

While the kick did not happen, we performed INITIAL_ALIGNMENT. During the initial alignment, we found that POP90 (LSC-POP_PDA2_RF90_I_MON) reached less than 0.8 even though PRMI ADSs were engaged (fig.3). Compared with the PRMI alignment 40 days ago (fig.4), IMC output also decreased to 95 % but POP related signals (PDA2_DC, RF90_I, and RF90_Q) decresed to less than 85 %. This indicates that something around POP decreased the POP light.

Then, we requested LSC_LOCK guardian to go to PRFPMI_RF_LOCKED at once. Sometimes, PR3 kick prevented the IFO from locking but HANDOVER from ALS to IR seems to be succeeded finally. On the other hand, Arm trans. powers seems to be low (X:~92, Y:82) Fig.4 shows the arm IR trans. powers when ARM ASCs with WFSs were engaged and PRMI ADCs were engaged. Also, PRFPMI lost the lock in ENGAGE_PRFPMI_ASC state, except for the cause by PR3 kick. 

We started to engage ASC one by one and found that PR3 Yaw control seems to be wrong. This control feedback seems to drift in one direction on engaging the control, even though the sign of the gain was flipped (fig.5).

By the way, we measured finesses of both arm cavities roughly. The results from 3 times averages are as below.  

X: 1353.6 +/- 5.4 (fig.6)
Y: 1249.9 +/- 6.6 (fig.7)

They seems to be consistent with the previous results in last month within each errors. So the power reduction seems to be not caused by finesse reduction.

We need more investigation. Today, I left the LSC_LOCK guardin in the ENGAGE_WFSDC state.