Now the lock with megatron is pretty easy. Really. It's very cool.
As we saw the oscillation of the YARM servo, we temporalily increased the gain of TRY filter by a factor of 2 (0.003->0.006). Also decreased the gain of YARM servo by the factor of 2 (1->0.5). This makes the servo gain reduced by a factor of 4 in total. This change seemed to come from the change of the ADC/DAC range.
We finally fixed the hi-gain pd transmission communications from Megatron to the c1lsc by tracking down the correct RFM memory location (which is unhelpfully labeled as a qpd channel in both losLinux and lsc40.m). The memory location is 0x11a1e0, and is refered to as qpdData.