As expected, the jitter feed forward that was tuned yesterday evening was no more good. Using the same method described before, I retuned it and found comparable performances as yesterday.

Addition:

The second plot shows how the coherence between DARM and DBB Q1Y changed over time. Initially there was no feed forward, then it was switched on at about t=1 hour. At t=8 hours I retuned it as explained above. Beware that the coherence color scale tops at 0.6 and not at 1. It looks like the subtraction was getting worse, but not too bad after all.