Something that isn't possible with our toolchain is to generate two separate distributions, because all grids are merged together in the end. For the CDF W-boson mass grids that would be beneficial and also for some of our top-pair production grids. We could distinguish distributions that need to be merged together and those which are separate by looking at the names of each histogram: merge grids together if their histograms have the same name, leave them alone if their names differ. Another concern is the metadata, for which we'll need as many files as there are different histograms.
Something that isn't possible with our toolchain is to generate two separate distributions, because all grids are merged together in the end. For the CDF W-boson mass grids that would be beneficial and also for some of our top-pair production grids. We could distinguish distributions that need to be merged together and those which are separate by looking at the names of each histogram: merge grids together if their histograms have the same name, leave them alone if their names differ. Another concern is the metadata, for which we'll need as many files as there are different histograms.