These are not addressing the block size scaling constraint. They are just referring to how fast nodes can verify transaction signatures. The former is the bandwidth elephant in the room and can't be scaled with existing consensus technology unless the mining is centralized on fiber connected servers.
In fairness there are many fiber connected homes now. Not everyone and not everywhere but it's become fairly common.
I can't even get reliable 3 Mbps download with 512 Kbps upload where I am. Many users will be on wireless devices.
Yes I as I said it isn't everywhere, but you don't need everywhere either, just "enough" places. I'd add that I've had 100-ish megabit wireless for quite a while too (and I sometimes use a location in a G7 country that just got upgraded to the ultra-speedy 3mb/384mb after many years stuck at 1.5/128!).
I don't disagree that massive blocks or fast blocks are problematic with a Bitcoin-style chain (nor that being functional over ham radio, mesh networks, etc. would be preferable) but I'm not sure that "increasing" centralization is inevitable with simple scaling either. It is a race between capacity and available connectivity. Six years ago when Bitcoin started the availability of even 10 megabit connections was far narrower. Gavin may be wrong overall but there is some merit in his approach too.