Somehow... this was causing a fluctuation in the readout time of ROC8... noted by looking at the timeout loop iterator (D.roc8rtry) vs. time (Entry$).... figure 1 (Red is 1 iteration... Black is more than 1 iteration)
After adjusting the threshold on the L1A discriminator (refreshes the L1A signal at the weldment) from -200mV to -300mV, the reflections do not appear to be an issue anymore. I did not see any missing gates for the few runs I took (with a DAQ random rate of ~2kHz).
I will keep an eye on it.
Figure 1