"As a matter of mathematics"

I'm taking another stab.

Let's call world population 1e10 (10 billion) to make the napkin math easy.

Total bitcoin supply is 2.1e15 sats, so let's say half of it is lost or locked up in deep cold storage that never transacts, again to make the napkin math easy, and work with 1e15 sats that is mostly transacting in lightning. For purposes of this thought experiment, we will posit that in our hyperbitcoinized future everyone will want self custody if they can afford it, even though present behavior might suggest otherwise (most now keep funds on exchanges and are afraid of self custody).

With no wealth inequality, that's 1e5 coins per human. 10k usd net worth at our hypothetical 10mm btcusd hyperbitcoinization scenario.

But there is always wealth inequality, so let's call it a lot of people with 1e2 to 1e3 sats (essentially no savings, $10-$100, actually probably in debt), and a few fat cat 1%ers, with hopefully a healthy middle class sandwiched in the middle with 1e5 coins, actually transacting on lightning. The very poor (1e2) will use custodial lightning and hopefully won't mind it too much because if they get rugged they can rebuild their savings buffer in days or weeks rather than years or decades. The striving poor (1e3-1e4) may use a combination of self-custodial lightning for savings, and bank-custody lightning for transacting. Occasionally the self-custodial lightning moves on chain for more secure savings.

You would like the middle class (1e5) using self-custodial lightning and on-chain for savings, without sweating too much about the cost. As I said in the other post, if fee rate rises to 10sat/byte (good for long term security budget), this won't really be possible and even the middle classes will tend to use custodial lightning or centralized clearing to save on costs.

Even with 1 sat/byte (>100 sat transactions), this is spending >0.1% of net worth for every channel ops (without shared utxos), maybe that is too much.

So that is my argument for why we may want to move towards a shared utxo model even if a case can be made for YAGNI. Maybe you aren't going to need it, but maybe you are going to need it. And soft forks are hard. So might as well allow it, if the downsides aren't too horrible. Will hopefully also make bitcoin more attractive to investors if it is clear that there is a well thought out scaling path that also keeps miners fat, happy and hashing, without enabling turing complete shitcoins.