However, wouldn’t stacking prints also lead to increase in grain? Obviously his samples were sharp, but I’m wondering what additional steps he took.
Stacking the prints would do that. But where grain is concerned, If I understand correctly, the software does more of a subtractive process: it compares each frame from, say, 5 sources and determines which elements are common to all, and which are unique to each. Any grain or dirt that appears in only 1 source is then rejected/eliminated, leaving only the stuff that is common with the rest of the sources as stuff to “keep.” With enough sources, this would theoretically leave only the image and grain from the original source (in this case, the negative) intact.