Regarding your point specifically, that is, computational limits, I think that these can be overcome eventually by using highly specialized "transaction validation" hardware...
Just like ASICs (or whatever) are used for mining coins today
So you think that someone is going to attempt to invent specialized processing units for the sole purpose of running nodes? You do realize that running a node only makes you constantly lose money, so there is zero incentive to attempt a large investment for such.
First of all, I don't think that we are ever going to get there, i.e. to 100Gb blocks, lol. As I said it a few times already (though not in this thread, thus you might have missed that), without changing the blockchain paradigm (e.g. from a flat blockchain to a multilayered blockchain), Bitcoin is pretty much doomed, and what is most scaring, we might be already in or quickly approaching this final stage. Having said that, I'm inclined to think that the very existence of specialized hardware designed specifically for mining Bitcoin proves my point. So, yes, someone would have attempted to design such processing units, to get a competitive edge over the rest of the pack, provided it could ever come to that in the first place...
And such units would have been designed long before confirmation of blocks became impossible without them