I took an open loop gain measurement of CSOFT P today, using Craig's unbiased method. The measurement is rough, but it gave me enough of a picture to suggest that increasing the gain might help. I have changed the gain from 20 to 25. This has been SDFed in observe.snap. We should monitor whether we continue to see locklosses from CSOFT P to determine if this was enough of a gain increase, or if the problem was from something else entirely. I'll post the measurement results compared to a model as soon as I have time to do the processing.