I'll just point out the distinction between "the code doesn't require it" and "this is safe". They aren't the same thing, but one would suppose the 3 psf in the code wouldn't be there over the strenuous objections of the engineering community (or, perhaps, it would), but there are situations where 3 psf wouldn't be safe, probably involving flat roofs that are deficient for slope to start with (courtesy of old codes that allowed "sloped to drain" rather than a specific slope, i.e. my kitchen table, while flat, is "sloped to drain" as when I pour a gallon of water on the table, besides a lot of yelling from my family, most of the water ends up on the floor so it's "sloped to drain", right? That's how it seemed to be interpreted. "perfectly flat" and increasing the dead load increases the chances of a ponding hazard or collapse and the definition of "susceptible bay" in the IBC isn't all that well connected to any charging language.
Steep sloped roofs you'd likely not encounter this situation. To me the 3 psf has an origin from shingles, and most shingle manufacturers don't (explicitly) allow slapping their shingles on old shingles now, anyway.
I do a lot of construction defect work, so ...... I tend to look at these things more holistically. And with a "what's the intent/origin of the provision" perspective. The NRCA does a fine job of fussing over covering secondary roof drains, or the need to add them, but the slope/ponding interaction, that seems to be a bit of a blind spot for them, as well as working on code slope requirements without noticing that the manufacturer's installation instructions also have slope requirements (as a backup).