Summary:

Y ARM IR QPDB on TMS, which is used for CHARD and CSOFT and also for estimating recycling gain, is severly clipped. This makes both the ASC itself as well as the power recycling gain estimate fishy (LSC-TR_? comes from QPDB).

Fix it (or try fixing it) before using such a fishy signal for ASC.

Details:

Attached is the trend of one lock where Evan increased the power from 3W to 16W. The alignment of the cavity was changing during this due to power up operation.

If you see the IR QPD SUM on TMS (middle panel),  X_A, X_B and Y_A are all sort of proportional to each other, but Y_B is not, so it's apparent that it's clipping.

If you make B/A ratio for X arm it's dead flat, but that's definitely not the case for Y arm (top panel). The ratio for Y should be about the same as that of X, but Y ratio is always smaller and it gets worse as the power goes up. 

Also, at around t=370 sec, there's a kink in YB that is not related to the power kink. Though the clipping was definitely going on at t=300 and probably earlier, it seems like the clipping got worse at t=370.