Sheila found this morning that the ETMY ISI seemed to have a large peak at 1.5hz, which has in the past been indicative of a failing L4C. Most of the BSCs are already running blends that bypass the horizontal L4Cs because, except for ITMX, all BSCs have a flaky H1 L4C. That seems to be the case here with ETMY, when I switched the St1 RX/RY blends to CPS/T240 blends the the 1.5hz motion immediately settled down.

The best way I've found to see which sensor is failing when this happens is to take the L4Cs out of loop and look at the local to local L4C to T240 transfer functions (first attached plot). This gives a measurement of the L4Cs sensor response, and in this case it looks like the V3 L4C (light blue traces) has now become an almost 2hz instrument. The V1 and V2 sensors are normal, what we would expect for an L4C. If you look at these tfs with the L4Cs in loop, you can't measure this  change in response. I haven't had a chance to try because of commissioning, but I suspect you also can't see this with the isolation loops off, or during a driven transfer function. The sensor has to be in a low acceleration state.

Just looking at the blend complementarity with 1 good L4C and 1 L4C with it's pendulum frequency moved to 1.9hz explains what is happening. The sum of all 3 parts of the blend multiplied by their respective sensor responses should be 1, but having mismatched sensor responses in the L4Cs causes a deep notch in the supersensor between 1 and 2hz, which I can easily imagine does bad things to the quality of the feedback controls, second image.

It's possible we might "fix" this sensor by giving the ISI a kick, but the L4C would probably eventually return to this bad state. Or we could try adding a filter to compensate for the sensor changed response, but it seem like this sensors mode has probably been drifting around for a few weeks at least. Third attached image show trends of log blrms motion for the St1 RX/RY over the last 30 days, typically these seem to sit at around -1, when the ISI is healthy, but it goes up to 1-2 when the L4C is misbehaving.