Following an e-mail exchange between Jeff K and Brett Shapiro about alog 37847, Daniel, Thomas Vo and I tried to follow Brett suggestion to swing ITMY and see if we could reverse the hysterisis.

First we took the ISI to isolated damped, then set the top mass damping gains to 0.  Daniel applied a 0.3 Hz excitation to M0 OPTICALIGN_P until the top mass was swinging with an amplitude of more than 200 urad according to the test mass osems (the sift we saw was about 600 urad).  He slowly ramped this down over 5 minutes, we let the suspension swing for a few more minutes.  We don't see any shift in the pitch of the optic before and after this test.

The second attachment shows the same channels during the EQ. The motion durring the EQ was much larger than what we applied with this test, so it might be that we need to try swinging harder.

Here is the suggestion from Brett's e-mail:

If you imagine that the hysteresis is coming from dislocations in the wires all moving to one side of the
wire or the other (front or back), then letting the quad swing and ring down slowly
causes the dislocations to spread out evenly, washing the hysteresis away. But if
the quad tips a lot, but rings down very fast, like from an earthquake causing it to
bang hard into the stops, then the dislocations may pile up on one side, causing a
static pitch. So the hysteresis plot I attached is really a worst case.

A simple thing to do to check if this hysteresis thing is the culprit is to get the
pendulum swinging in pitch, with an amplitude at least as large as the offset you
see (if possible). Then let it ring down without damping. If the offset goes away or
gets smaller, then this was it. If not, then it must be something else.