To increase geometry, why wouldn't we just crank up the subdivision surface mod before applying it, instead of lightly inflating and smoothing after the fact?
The light inflation isn't so much for increasing geometry density but rather to dyntopo-ify the wireframe - make the topology as consistent as possible. I find that when some areas are dyntopo and other areas are subdivisionally-smoothed, it's annoyingly noticeable. So there is some preference in there. But in general when sculpting with dyntopo, I never want more geometry than is necessary.
