Craig showed that the ASC signals, including CHARD_Y, get worse during thermalization. In particular, noise above 10 Hz gets much larger.

I looked at bit better at the increased noise in CHARD_Y during thermalization:

1. it's visible in both REFL_A_45 and REFL_B_45, the two signals that are used to compose CHARD_Y
2. the same increase is visible in the longitudinal diodes LSC-REFL_A_RF45_I and Q 
3. the coherence between the WFS and the LSC inphase signal increases with thermalization, so they are seeing the same noise
4. the transfer function of WFS / LSC is suggestive of a close loop response of a loop with a broad gain peaking at ~5 Hz