I had a look at three of the locklosses that happened durring ENGAGE_ASC over the weekend. 

In all three cases, we were hitting the soft limits on SRC2, and SRC1 Y in the seconds before the locklosses.  The error signals of the AS centering loops, DC3+4, are pretty far from 0 at the same times, and there is an oscillation in DHARD at around 4Hz.

One idea to try to untangle what is happening here is to turn off some of the soft limiters, I have edited the guardian for the moment to remove soft limiters on SRC1+ SRC2, as well as INP1.