Essentially, we need a way to reduce the number of transactions that recorded on-chain and increase the number of actual transactions. That's the basics of a scaling.

I think it might be lightning on drive chains, the more I thing on it. I don't exactly want an update to the consensus method, but I feel something has to be done.

As for LN, it makes two on-chain transactions and an arbitrary number of off-chain between two parties. Routing makes this expandable, but a network is still needed to do the routing, which becomes hard to navigate as we inevitably go into the millions of nodes territory, and getting to the millions of channels territory means millions of transactions to establish and maintain those channels. On top of it, this means non-custodial for the vast majority of people.

If people within drive chains could setup lightning channels between themselves, having your own drive chain lightning channels might actually be feasible for the average person as we venture into the hundreds of millions and eventually billions of users territory. Obviously, we would have cross-chain nodes that would route across drivechains, so that becomes an engineering problem, not a philosophical one.

10k chains × 100k users per chain = 1B users.

Every medium sized city could have its own drivechain! That could be quite feasible. (No, I do not mean government run, just that a chain could be associated with a city.)

There's also the idea of nested drive chains. I could imagine 1k drive chains with 100 sub-chains and 10k users apiece to get to that billion person mark.