Yes, I was referring to this old study. Of course it would be nice to see a newer opinion on that, if you know any newer one I'd be interested in reading it. Googling has mainly brought content from the olden blocksize debate tymes (2017/18).
I interpret that the UTXO set growth should be lead to even higher memory usage (if higher as estimated back then). The compact blocks instead should be mainy being affecting propagation, and not so much verification, because the transactions have to be verified anyway against the UTXO set only once I guess (I don't know if they were verified twice before compact blocks were introduced?).
At least I think the requirements should stay in the same order of magnitude, but of course I may be wrong if there were some outstanding discoveries in the last years.
Thanks for the link. I however challenge one of the first arguments made up there:
Lightning channels require at least 2 on-chain transactions to set up. If fees even stay at their CURRENT prices (which, I predict they will increase as the demand for block space increases), then that is widely unaffordable for most people in the world. Sure, you could interact with the network through a third party, but that defeats the purpose of Bitcoin.
They forget sidechains and rollups - which are important topics as they seem to work well on Ethereum. Sidechains allow people to "onboard" completely off-chain (from the mainchain's point of view), and ideally they would not ever have to touch the mainchain. This is an important advantage regarding Lightning.
I already saw projects like BEVM, Nomic, tBTC/Keep Network (not to be confused with testnet, of course) and (upcoming) Stacks/sBTC which may have found a reasonably decentralized way to operate a sidechain (on a DPoS-style setting with a federation with up to 1000 participants, which operates in BFT style consensus, e.g. similar to Ethereum), although as I mention
here the big problem imo is currently the semi-centralized aspect of these chains with 50% premine of the tokens.