The problem is the way it contradicts your other argument, including "The d20 system tells me that I succeeded, but provided no help to the GM on what succeeding in this social set-up move actually meant.". Rolling d20+ bonus vs target number is fast and easy, but doesn't give you what you want.
Happily, IMO, it doesn't actually *get in the way* of producing rules for what you want: for example, you could just implement a dynamic DC now, of
5 - Chloe wrings her hands.
10 - Chloe wrings her hands as she tells you that the man is asleep, and she doesn't know what to do.
15 - Chloe wrings her hands as she tells you that the man is asleep because she drugged the nearby lake with a sleeping potion, and she doesn't know what to do.
20 - Chloe wrings her hands as she tells you that the man (and also that deer over there) is asleep because she drugged the nearby lake with a sleeping potion, and she doesn't know what to do. If you can wake him, she promises to pour the antidote into the lake.
Actually, if you look back, I was originally talking about "persuasion", which, as most any 5-year-old can attempt it (some even successfully), hardly seems a "specialized skill".
My original argument was, getting 40 random unskilled people to make an argument is, on average, in 5e, more convincing than the most talented speaker alive.
Same with perception: keeping dozens of inattentive bannerman, porters, and torchbearers generally produces better results than having an eagle-eyed PC.
EDIT: and putting 40 average programmers on a task? That *definitely* won't produce code up to my quality standards, so I can definitively state from a position of experience that modeling programming represents an epic failure of Bounded Accuracy to model reality.