Something is really wrong but we haven't figured out what exactly that is.

As of now, IMC_LOCK guardian locks IMC automatically if we wait long enough. The problem as of now is that the FSS is hit hard when H1:IMC-REFL_SERVO_COMBOOST is enabled IF the residual low frequency component of the IMC lock error signal  (H1:IMC-I_OUT_DQ) is big. In such a case FSS is unlocked and sometimes FSS autolocker goes into a loop of state 2 to state 3 to state 2. (The problem was made worse by the JAC guardian's behavior to unlock itself whenever FSS is unlocked, and that the JAC auto locker somehow kept trying to lock to the fringe at a very low PZT voltage and immediately unlocked, but these JAC problems were solved now. See alog 91105.)

However, IMC doesn't lock with nominal gains. I had to considerably reduce MC2 gain: Overall gain was reduced by a factor of 0.035 (H1:SUS-MC2_M3_ISCINF_L_GAIN=0.035), and the M3 gain was further reduced by a factor of 0.2 (H1:SUS-MC2_M3_DRIVEALIGN_L2L_GAIN=0.2). M2 and M1 gain are nominal. (Probably the overall gain should be set upstream like   H1:IMC-MCL_GAIN or H1:IMC-L_GAIN, but that's not what I did.)

Once IMC is fully locked, IMC stayed locked unless we started changing gains too much and/or injecting huge signal. Also, fully locked, I can flip the COMBOOST switch OFF and back ON quickly without killing FSS.

We measured the IMC crossovers but they were all very wrong. 

On Monday we should measure the overll OLTF on the floor to see if the fast path is still good. If it is we can focus on diagnozing the MC2 actuation, otherwise there could be a deeper problem.