There is no doubt that a single forever remembered blockchain cannot handle every tip or coffee bought so offchain or better sidechains are needed. But that does not mean that BTC should not also be scaled in block size. With the price trending down discouraging new holders we need to welcome any and all real use. And approaching a txn limit does not help that.
I would not be at all surprised if Bitcoin were to become more in-demand if it DID start to cost something to use it. Psychologically, if someone is given something 'for free' than it is not valued very highly. As soon as it starts to cost something then it starts to be more highly prized.
Since most thinking persons are starting to catch on to the fact that it does not make sense to broadcast every cup of java purchase to every decentralized infrastructure provider (to hold forever) and Bitcoin's only realistic role is to be for higher importance transactions, having to pay at least to cost of network operation for transactions is, in my opinion, a positive and something we should be migrating towards. Probably by shortly after then next coinbase notch-down (roughly 8 years from the genesis block) we should get to that point. It's very much up-in-the-air about whether that will happen even if the block size stays at 1MB.
You have got the pricing of transaction backwards. If it is usful to transact, and the transactions are limited, it will have a price. If there is a nontrivial price, that means it is usful. What will happen when we reach the limit, is that the price of transactions will rise, and then, if the block size is increased, the price will fall again.
We obviously need to be able to serve lot of transactions, if we want bitcoin to be widely used. But costly transactions is not a definitive stop, it will only deter the least important transactions. USD10 is still cheap, and you can do a lot of mining for USD10 per transaction.