We are still far from the point (although it's years, not decades), but.. let's see what the hardware brings too in that time, OK?
Well, we've experienced quite a few bouts of elevated fees nowadays and probably is discouraging adoption as well. Not as far as you think.
Indeed, second layer channels do need on-chain operations. I knew that. But we don't know yet how big will be the need to open and close channels. Well thought channels will not need closing because they'll keep carrying transactions in both directions. Am I missing something there?
Not really. LN is more suitable for microtransactions due to the liquidity of the channels and you will at some point require on-chain transactions. It's simply impossible for us to function solely on LN, or at least without a fairly huge dependency on it. LN is really not the panacea to the problems that we're facing right now. Yes, it will help to alleviate the issue but it really doesn't do that much. If we want any significant adoption, then a block size increase is the most straightforward method.
There are a few threads, off the top of my head that I've discussed about LN and the need for on-chain transactions as well. I think that would be a better starting point.