Nobody said that propagation delays are the *only* obstacle. We said they are currently the *primary* obstacle, because currently you will run into the trouble at much lower tps.
That's what I understood here:
Surely you must agree that if there were no block propagation delays, there would be no centralization associated with scaling up. Blocks go instantly to everyone, and there are no problems with miners getting more than their share of the blocks.
I agree with you that it is quite likely that BTC is less scalable than the dollar and other centralized systems, but it is worth it to try and get as close as possible. Plus, you still can't completely rule out the chance that someone will find a way to distribute the workload and not just distribute trust among the many nodes.
That's were we disagree, I think. As close as possible may be sacrificing too much decentralization.
As said, this doesn't mean that the proposal isn't interesting for increasing scalability to a "decentralized-enough level", whatever that is.
Plus, you still can't completely rule out the chance that someone will find a way to distribute the workload and not just distribute trust among the many nodes.
Distributing the validation workload would be ideal. But I think that such a system (if possible) would be very different from bitcoin in several ways.
You would need some zero knowledge proof (ZKP) system so that the rest of the nodes could trust a single validator. I'm also very interested in that kind of proposals.