I'm not sure I've ever heard this view expressed by serious, competent network engineers. I have heard it a lot from the home hobbyist though, but I'm not sure how much that demographic matters in the grand scheme of things.
Additionally, like a sibling comment notes, a home hobbyist has full control over at least half, often more, of their addresses and can easily choose addresses for their network that are as short or shorter and easier to remember and organize vs a v4 network where you have no letters to work with much more strict subnet size rules, etc.
IPv6 is a dream for home hobbyists! The complaining from them about “unmemorable” addresses just makes no sense.
Well, the non-trivial percentage of large orgs that have literally run out of RFC 1918 space would disagree.
But yes, you're right. There's a weird Stockholm syndrome thing some people have with NAT.
It'd be hard to have so many devices that even in 10.0.0.0/8, you run into a need to have letters as part of the network addresses.
My home network is larger than most and I while I use multiple subnets for fun, I could it all of it into a single /24.
Thus ULA is a must on the inside, and DynDNS is still required for anything internet facing.
In the real world, people who design and operate large networks are the very same people who staffed the working groups who designed IPv6. It's their design.
E.g.: If I want to deploy a cloud VPC (or vNET), then I have to go find "the guy with the spreadsheet" and peel off a tiny(!) private IPv4 address space. If he's away from his desk or on holidays, my 1-minute automation script will now take 1-10 working days until he's back and responding to requests. With IPv6 this just disappears as a bottleneck.
This is not the kind of glowing endorsement you think it is, if you're expecting your technology to see widespread adoption