Banking = promise where you have no idea if the agreement will be honored.
Payment channel = you know for a fact that the channel can be closed, settling the fund transfer
Imagine you're a programmer on an accounting system - like Sage or Dynamics or something like that.
You agonise for years over the perfect security systems, audit trails, corporate transparency and compliance support logic.
Then the end user just bins today's datafile and restores yesterday's blowing all your beautiful logic to smithereens in a an otherwise obvious action that simply did not belong to the domain of said 'security logic'.
The ability to do that or not do it is the difference between a bearer token and promissory note for such.
Any honest apologist for bitcoin would admit what the real priority is which is leave the protocol untouched, and not to improve its monetary properties natively.
Well, there are tradeoffs involved and you have to pick which monetary properties you'll keep.
If the pace of onchain scaling growth exceeds the pace where users can catch up in terms of hardware and connectivity, at that point you are leaving behind the property of decentralization and increasingly become centralized.
But we are not here to make paypal #2, are we? We already have a centralized paypal. The question is how to make a decentralized one. And if users want many tx/s, if they want microtxs, if they want instant txs, if they want decentralization, if they want lower costs, etc etc, all these can't be achieved with onchain scaling right now.
It will be possible in the future though because tx needs are finite. As a species we only conduct so many txs per day. With a client/server model that has no redundant/p2p inefficiencies, it's pretty easy to cover them. When technology evolves, it will eventually be easy to cover them even with a blockchain p2p approach instead of using workarounds like payment channels. We'll see how it goes although it might take decades.