I've written a new guardian state in ISC_LOCK that will turn on all ASC loops at low bandwidth, wait for convergence of them all, and then increase the bandwidths of the loops.  I have tried this once by hand a day or so ago, but never with the guardian.  This should, however, solve the engage asc locklosses that we have when some of the early loops (eg. PRC2, INP1) pull the IFO alignment somewhere, and the other loops (eg. SRC1 and SRC2) aren't able to follow.  This pulls the recycling gain down, but then when the next loops are engaged all of the buildups get better. 

Since all of the full IFO ASC is in one state, if you were to get to PREP_ASC_FOR_FULL_IFO and just request something at or higher than PREP_DC_READOUT_TRANSITION, the guardian will take you through the new ENGAGE_ASC_FOR_FULL_IFO.  However, if you get to PREP and select one of the old ASC states, like ENGAGE_REFL_POP_WFS, then the guardian will take you through the old set of ASC engagement steps. The attached screenshot of the guardian graph hopefully makes this more clear.