Good question. This is the idea that ultimately on-chain will become more of a settlement layer than a transaction layer, right?

Wouldn't this balance out on it's own, though? As demand for on-chain transfers falls it will become more affordable and people will do more of it.

Also, for the next 100+ years, miners will still earn bitcoin. That should be enough incentive, because mining costs are dynamic and market mechanisms will adjust them accordingly.

My view is If fees are low it encourages more self custody and more on-chain if fees are high it just moves certain economic transactions from going on-chain for the time being, which will likely be consolidated on-chain when capital needs to be stored long term, because there's no substitute for a trusted on-chain UTXO

On my mind is this idea of a "unilateral exit box". So in lightning we have the balance of my side of the channel and the balance of your side of the channel, and I think we should have a balance for miners that increases with each channel update (so its paid incrementally as part of ln routing fees).

This construct would SUCK without Eltoo/LNSymmetry, because it would mean your channel partner just sent an old state update using your aggregated mining fees and you have to pay your own mining fee by yourself to get a justice transaction. With Eltoo/LNSymmetry, you just replace their transaction with your updated state and use the same mining fees.

This construct scales out to when a bunch of people are all updating the same channel (multi-party channel) and when multi-party channels are issuing timeout tree vtxos to casual users.

And that's how I think we'll scale lightning with unilateral exit while paying extremely high mining fees in a pain minimized way.