regional plots, such as those done using locuszoom (http://locuszoom.org/) can be very useful to explore GWAS signals. I think all the necessary tools are already available in Hail (such as filter_intervals and ld_matrix).
One would feed a list of variants, and a distance from aforementioned variants, and a function would create a structure that could be plotted. It would have to contain all variants within distance D of the variant of interest (filter_intervals should do nicely), as well as the ld between each variant in that interval and the variant of interest (so in this case, something lighter than ld_matrix as we basically need only one row/column from that matrix). I think the recombination rate and the genes can be loaded in the global annotation (Or should this be in the variants annotation?), and therefore made available for inclusion in the output structure.
Do you think that would be possible?