Thank you so much for summarizing and bringing up great questions! I’ll try to provide my perspective here, hoping that other community members will weigh in as well.
I don’t have the data readily available at hand (maybe @irina can help here), but I’m definite that these two are very distinct sets. You can get some stETH metrics here: Dune.
Judging from Etherscan data, about half of total stETH supply is held by three protocols: AAVE (31.5%), Curve (12%), and Maker (5.5% via wstETH token). The rest of stETH is distributed in a much more diverse way, and for the three protocols, it would be both very damaging reputation-wise and non-trivial technically (they’ll need to vote for a protocol upgrade) to use stETH they hold in a Lido veto voting.
The fact that Lido is a multi-chain service is one of the causes of the principal-agent problem this proposal tries to address: due to the Lido governance having power on other chains, LDO holders (the agent) may have incentives that are misaligned with Ethereum stakers (the principal).
While I beleive that the Ethereum liquid staking protocol is the central service Lido provides, and will remain so in a foreseeable future (and, as far as I can tell, the whole team beleives the same), the mere possibility of having non-aligned incentives justifies introduction of a safety mechanism like the one proposed here.
That said, multi-chain governance is just one cause of the principal-agent problem which is present even in mono-chain setups.
While I think this is a good outcome for Ethereum given that Lido will be able to eventually ossify its core contracts, there are other optimal outcomes, e.g.: most stake being taken by small independent entities each holding 1-2% of stake at most; an oligopoly of staking solutions with varying degree of centralization, each taking no more than 10-15% of stake; 3 largest solutions taking 90% stake, but each a with sufficiently decentralized validator set and network-aligned governance. The question is whether these other optimal states are reachable, and the opinions here vary greatly accross the Ethereum community.
In my opinion, the most probable outcome is the largest service taking the majority of stake (say, 50-60%), with the next 30-40% being distributed between 3-5 large solutions. In this case, it’s critical for the largest service, be it Lido or some other solution, to provide a robust and diverse validator set and to be aligned with stakers.
Integration of liquid staking mechanics into the base protocol should result in a less complex and more liquid setup. But that doesn’t mean this setup will be more decentralized—I’d argue it will be quite the opposite.
Given that ETH is not and should not be a governance token, the stake distribution mechanism has to be non-opinionated and completely based on data available onchain, which basically means that any entity would be able to register as many validators as they can economically and operationally afford, no matter whether these validators are distributed jurisdictionally and geographically, whether mono-client or whitelabel setup is used, and so on.
This setup, combined with the effect of economies of scale, will result in the majority of stake being anonymously taken by large entities (and/or entities running whitelabel nodes) that have no incentives or obligations to decentralize and generally act in the best interest of the network and stakers.
One might think about some mechanism that allows bringing the meatspace data onchain to introduce decentralization incentives (and we’re actually researching this as part of the Lido’s upcoming permissionless node operator subset), but it’s a very complex research that will require multiple iterations over several years. Moreover, integrating it into the consensus rules will make the latter significantly more complex and fragile, so I’d argue that the best place for the validator set selection mechanics is still the more flexible execution layer.