I had a look at the SRCL coupling to DARM during last night lock. Indeed it is non stationary, and the transfer function changes shape significantly. It appears, as already seen, that there is a low frequency component which is more or less constant, plus a high frequency component with a quickly varying gain. See the first plot for an animation of the TF from SRCL output to (calibrated) DARM.

I analyzed the coupling at 40 Hz and at 300 Hz. For the low frequency, the fluctuations are mainly in the transfer function gain, which is on average about 3e-11 (in some units) and fluctuates between 2 and 8e-11. For the high frequency, it seems that the phase as well as the amplitude changes. So I analyzed the real and imaginary parts of the TF separately.

The attached plots show that the coupling fluctuations are very well correlated with angular motions. In particular H1:ASC-AS_A_RF36_I_PIT_OUT_DQ is the winner. This signal has the same content as H1:ASC-AS_B_RF36_I_PIT_OUT_DQ.

So my conclusions is that the non stationarity is still dominated by angular motions in the SRC. Indeed, we didn't switch on any ASC boosts on SR1/SR2. We have to improve SRC angular stability. In particular SRC1_PITCH