The white paper is always brought up. And Satoshi had not anticipated ASICS (who knows if that later Satoshi email is real where "he" talks about exactly that) and that changed EVERYTHING. It made the idea of the "validation node" important. Vitally important. We discovered by this tech that between hashing, and consensus validation the latter *might* be more important, though they go deeply hand in hand.
Many would have been happy with cautious size increases. There's no real justification behind the 1MB (as would be expected from it being implemented as an anti-DOS measure when blocks were tiny) and the theoretical maximum while still maintaining your goals is much higher than that.
I'd advise not getting wrapped up in Raspberry Pis. They are cool devices and I own many but they are not optimal for this kind of service by any means. They are radically underpowered and even within the realms of SBCs, there are more powerful options available. It's not a binary choice between cell phone processors from the early 2010s and supercomputers in datacenters.
I also disagree with the usefulness of "validation nodes". They don't have zero value but they're not as useful as people think and they're certainly not worth crippling the network for just so people can run them on sub-par hardware. If your security is worth $40 per transaction, it's certainly worth more than a $40 computer.