You are positing an extreme in which uke detaches and falls down. Did I ever say that?
I am positing a moderate training path in which uke stays attached and gives feedback, body to body, appropriate to the level of the partner. Incrementally correct movement by nage shows as a reaction in uke. Uke doesn't try to make nage fail - uke's body guides nage to correct movement.
This makes sense to me. There needs to be some kind of feedback for any person or animal to learn something. No one, either humans or animals, learns behaviour patterns by seeing the same response regardless of what they do. That goes both ways -- all success or all failure, either way there's no information in that, and consequently no learning. I don't know how to express it without going mathy, but you need at least two values to encode information! A computer that had only 1s or only 0s wouldn't be a computer, it would be a plastic and metal box.
I suppose that doesn't actually mean uke has to go down, but you have to at least be able to see somehow which of your attempts were somewhat better than your other attempts. It's like playing a game of 'hot' and 'cold' -- you're unlikely to get anywhere if don't know if you're getting closer or further, if you aren't told anything until you're actually 'there'.
I also want to point out that switching techniques to one that's easier in a given situation is only easy if you know how to do it. When I was 4th kyu I found that extremely difficult! You don't automatically see the opportunities, let alone see which one will work best. I totally agree that it doesn't make sense to always do that, but it's still an actual skill and takes work to learn and doesn't necessarily come easily or automatically to everyone.