It is clear to me that penalizing assumptions rather than "stuff" has worked out well in the past. It isn't clear to me that we shouldn't penalize stuff (or in this case, universes) at all. The fewest-assumptions version of MW seems to lead to such an explosion of "stuff" that there might well (in our optimal version of Occam's razor) exist some term that doesn't show up in evaluating nebulae, but dominates in considering MW. That said, applying a meta-Occam's-Razor for now should leave us with the focus on assumptions, and thus I think with MW.