The problem with that quote is that if "the rest" of the client nodes also can't validate, then Bitcoin is centralized and completely pointless. Mining is already effectively centralized, the only check and balance against collusion of miners is a robust and engaged community of users running full nodes.

The client nodes have to validate in addition to only doing transactions. Satoshi doesn't say that in his comment. But the faster the block size growth, the faster it gets to "every Usenet user runs their own NNTP server", and the fewer and fewer run full validating nodes.

His comment is self-contradictory.

It’s a complex socio-economic-technical system, which probably can’t be perfectly quantified. Same as with the weather or the larger economy. We can understand it to some degree, but lack of perfectly predictive models does not invalidate these concerns, as you imply.

“Everyone must run a full node” is aspirational but not realistic. It’s nevertheless extremely valuable to continue working on ways of reducing the expense of running full nodes. MimbleWimble, Coda and others are doing a good job of exploring that problem space, as are some projects in Bitcoin that may take longer deploy.

When HN first started discussing Bitcoin almost a decade ago, the smartest skeptics here main objection was the obvious one that a distributed database where all the data is replicated across every node and which grows infinitely is likely not viable. They were right then and right now, it’s a hard problem and arguably the main existential risk to Bitcoin.

Throwing caution to wind so Bitcoin can have fast payments Now at the expense of failing at sound money later is short-sighted and irresponsible.