For many distributions of number of events, having part of it plotted in linear scale (top part, high-stat; bulk) and part in log scale (bottom part; low-stat; tails) may be useful in assessing agreement without the need to have separate lin and log plots.
Before going about doing this the wrong way (I thought of drawing things twice with overlapping pads in a canvas, which is a pretty ugly solution,) I was hoping that ROOT experts could lay out a roadmap on how this could be done in a more elegant and re-usable way.