Flashbots

A non-aggressing order (I'll use post-only limit as an example, but as you probably know there are a zillion kinds of PoL and IOC and flags and stuff that meet this criteria) strictly makes the book deeper, or narrower, or both.

A pension fund (or rather the agency execution-type firm that they'd be crazy not to employ for block trades) wants deep books and narrow spreads: if they were in the business of collecting maker rebates then they'd be market makers, which would make them terrible pension fund managers.

If their agency execution firm is able to shave a bit of slippage/friction off by leaving some passive orders in the book I suppose they might have some luck with that, but there's no way that punishing retail traders with wider spreads on the off chance that an agency execution firm has to have a socially-approved client is a net win under any definition of socially positive. Execution cost for non-advanced actors is wide spreads, shallow insides, and high fees: strict market makers improve all of that.

People who make arguments like this one are usually day traders with no business using limit orders. I'm not saying that's what you are, but that's usually where you hear it: certainly to the extent that this argument has an intellectual pedigree other than that awful Lewis book, that's it.

The moment you have to stake 32 ETH to be a validator, it was already made highly inaccessible for most.

As soon as that node ends up running on AWS rather than a user’s typical laptop as suggested by Vitalik and his friends, then that was another decentralization failure there and there is no self-hosting at home.

Finally, when the merge was finalized, there where only two addresses that ended up holding more than 48% of Ethereum that is used for the PoS mechanism to work. Another centralization risk feared that happened in reality.

So, that looks to me like a failure of decentralization and a more centralized system that as expected; made worse with Proof of Stake.

Not quite. It says this "may be altruistic". "May" does a lot of heavy lifting there.

From an outsider's perspective, a more likely explanation is that the funding is a subsidy. The VC having a stake in this space would have invested in a lot of other companies too. If their funding for Flashbots is notably less than the aggregate expected value of their total investment in the ecosystem at large, and if without Flashbots that expected value would drop by more than the funding amount, then the subsidy would be a net positive investment for that VC.

In other words, they may have seen Flashbots as an ongoing cost of doing business.

And if they had indeed achieved staying centralisation, there's always the access fee structure to hike up.

Mev-boost, a piece of software that consensus clients delegate their block building to, is currently being used on mainnet. The difference is that there can be multiple block builder _relays_, such as Flashbots, bloXroute, Manifold Finance, and Blocknative.

These relays effectively accept bundles from MEV searchers, which are then sent to the validator running mev-boost as they propose their block.

https://boost.flashbots.net/

Somewhat counterintuitively, PoW is inherently centralized. Its properties that make it viable at all also make it easy to become controlled by small number of big actors. This is hardly specific to Ethereum.