I did the calculation a while back and when bitcoin was only a few thousand dollars, if there were no coins bieng mined there would have to be the equivalent of $60 fee per transaction for the miners to make as much as they were making for the coins they mined. Now with bitcoin at the price it currently is, it would cost hundreds of dollars per transaction. Do we really believe that bitcoin could survive on just fees alone?
Since the original design didn't really have a blocksize limit, it's likely that the plan was to just have more users transacting on-chain and the fees spread more evenly between them. That way, fees would be healthy and fair for both miners and users. But with mining becoming so specialised in the ridiculously short space of time it did, that effectively negated any practical chance of normal, everyday people running what the whitepaper (arguably mistakenly) assumed would be a "node". However, It's important to raise the distinction that every time the whitepaper uses the phrase "nodes" it actually means what we now refer to as "miners".
Instead, we ended up with this awkward middle class, who unfortunately get none of the advantage of block rewards and all of the burden of maintaining a full copy of the ledger. Those are what we now think of as full nodes that form the backbone of the network and miners are something else. The original concept only had miners and SPV users. There were no full nodes that didn't mine. So in a way, it could be seen as slightly perverse that those who run full nodes claim to have the most power and influence over the network when they weren't even a part of the design, but it sadly couldn't be avoided after mining became more centralised that it ought to be. So, for better or worse, we do now need full non-mining nodes for the sake of decentralisation and we can't ignore the importance of that.
If it weren't for that hiccup, there would be less of an issue and on-chain scaling wouldn't be so contentious. More throughput would mean a greater quantity of fees being paid, so the individual fees wouldn't need to be so high. But we can't increase throughput significantly without decimating the full node middle class who we rely on to preserve decentralisation. Basically, it all went a bit off-script due to unforeseen technological advancements (namely ASICs) and now we're making the best of a less-than-ideal situation. Perhaps it does raise questions over Bitcoin's longevity, but it's likely we'll find a solution later down the line.