ISS has been oscillating at around 1.x kHz since 6:30 AM or so (first attachment). This happens once in a while and no fundamental cause has been identified yet.

Anyway, I disabled ISS and turned it on again for recovery, and in doing so I found another familiar failure mode caused by the diffraction power set too small (H1:PSL-ISS_REFSIGNAL was set -1.7).

Whenever AOM hits the bottom, ISS receives a huge kick and the AOM output goes very high, and for whatever reason it comes down very slowly and eventually settles. I wonder if this slow impulse response is an intended behavior or not. Anyway, whenever you enable ISS this happens at least once, it seems. If the diffraction power is set too small, the AOM will hit the bottom again, and the process repeats itself (second attachment). If you start with a higher diffraction and then slowly bring it down, it will work for a while until a big intensity glitch hits.

The third attachment shows the ISS first loop with barely large enough diffraction (H1:PSL-ISS_REFSIGNAL= -1.64, which resulted in 10+ percent of mean diffraction) for the intensity noise as of now. The servo was turned on at t=-190 sec or so on the plot. You can see how frequently the servo gets close to bottoming out.

I increased the diffraction further (fourth attachment, H1:PSL-ISS_REFSIGNAL=-1.6) and adjusted the offset.