While watching the lock acquisition, I was surprised at how long we were in the state MOVE_SPOTS before the A2L gains are actually changed. Looking at the code, it is because there are some sleep statements in there for initializing the TMS centering servo that don't need to be in there.
Recall that we use a servo (that Craig implemented over the October break) to move the TMSs such that we are centered on the IR trans QPDs, and maintain that centering as we engage ASC and change the IFO alignment. The first time the servo is used in ENGAGE_SOFT_LOOPS, we are careful to zero the offsets of the TMS TEST filter banks carefully, if they were somehow left on (they get turned off in DOWN). Then a cds.servo is used, and when we leave the state we leave the TMS in its new pointing.
The servos are started up again when we go to MOVE_SPOTS. But, we seem to be resetting the pointing of the TMS before starting the servos up. The way the reset code was written, it waits 30 seconds to ramp off each TMS offset, but it does it in series, for a total of 2 minutes of sleeping. We shouldn't be resetting the offsets here, because that erases the centering that was done earlier, so I commented out this whole section which also removed the sleeps.
It's possible that the IFO liked this 2 min wait to get some time for convergence and thermalization after going up to 10W. If there are problems at this state, please uncomment lines 3611-3621 in ISC_LOCK to put these sleeps back in (if this is needed, I'll fix them on ~Tuesday so that they are timers and not sleeps, since we're not checking for lockloss during each of these 30 sec sleeps).
Separately, we should probably be resetting the A2L gains to their centered locations in PREP_FOR_LOCKING, rather than waiting until PREP_ASC_FOR_FULL_IFO, since we engage DHARD before prep_asc_for_full_IFO. Seems to not be a big deal, but we are jerking that around on poor DHARD when we set them to their centered values just before turning on the dither lines, so we're changing the amount of DHARD noise injected to DARM.
The lines to uncomment are now 3616-3626 if needed, since I added a few lines in PREP_FOR_LOCKING to center the A2L gains (see alog 54950).
We lost lock once in move spots today (after Jim did an initial alignment). I serached on the lockloss page for locklosses at move spots (506). We've had two lcoklosses here since this change was made, and we had 3 locklosses from move spots in January. I rearranged the order of these guardian states on Jan1st, 54219, which is how some of these pauses got in here. Since we've only had two locklosses, and we made it though for now, I'm not going to make a change, but if we keep having locklosses here it might be that we do need to wait for the TMS servo.