The RPD-OR behaved linearly over the range of a moderate intensity scan, but the slope of that line was significantly reduced after raising E-counter discriminator thresholds (subsequent to careful gain balancing; the quality of this balancing can be seen in ADC distributions). The effect of the threshold increase is confirmed in ADC plots before and after the change, in which it can be seen that the minimum ionizing peaks have all but disappeared. [Looking at the same plots over the high intensity scan shows a return of these peaks, an indication of accidentals inside the ADC gate.] The positive effect is also seen in the linearity of the RPD-OR over the same high-intensity scan (up to 175 nA).
G0 was dying through much of the early part of the run, and was replaced. It and E0 and E1 were the exceptions to the quality gain-balancing.
In case we run with the RPDs again next year, it's worth trying to
understand E vs 
E to see if anything can be gained from
this information.