Thank you for engaging @vbuterin. I’m going to describe a few different attacks that I see as being possible under PBS/MEVA. I’ll do this across several posts, mostly to give myself time to write them up. Finally I’ll respond to your insightful post on consensus, which may take me longer.
Attack 1: Secondary Censorship Market
As you rightly point out, it becomes increasingly expensive for the dominant extractor to censor perpetually for their own ends, in most cases prohibitively so.
A participant must coincidentally be one of the best at extracting MEV, as well as having a good reason and deep pockets if they are to censor other participants continuously.
But the requirement for a coincidence of this kind is trivially solved by markets, and this is what I see happening.
The dominant extractor runs a secondary market allowing users to bid to exclude transactions from a block. It’s like an inverted PGA. You send the hash of someone else’s transaction that you don’t want to be included and a bribe. The dominant extractor will only consider your bribe if it is more than the gas (and MEV) that they would have received for inclusion, therefore they will be guaranteed to profit from it.
But the situation is worse than this because the dominant extractor can also offer a protection service allowing users to send the hash of a transaction that they want this censorship cancelled for. This will only be considered if it is more than the highest censorship bid for this transaction.
If the dominant extractor runs a lit market, users will be able to see if their transactions are being censored and can outbid their attackers to be included. This will lead to a bidding war with two losers (the users) and one very wealthy and ever more dominant winner (the dominant extractor).
If they run a dark market, users will have to guess whether their transaction might be censored or not and will often pay for protection they don’t need.
The dominant extractor will run whichever market type (lit or dark) is the most profitable for them, and possibly both.
In this way, many censoring participants can target individual transactions for only as long as they require censorship, but the dominant extractor can remain in power indefinitely simply by being the best at running this censorship market and pretty good at extracting MEV.
It works because the censorship/protection market acts as an efficient way of extracting private MEV (eg: censoring competitor transactions), as well as public MEV (eg: DEX arbs etc), a distinction I will discuss later.
Crucially, only those extracting this extra value by selling censorship/protection as a service will be able to afford to win blocks.
They will have won dominance over blockchain content by necessarily being the most exploitative. As well as the centralization that comes from this additional censorship profit, any network effect around their censorship market will act to further centralize around them.
Before long Flashbots will need to decide whether they are willing to cross the line and offer transaction censorship/censorship protection services to customers (something they weren’t with reorg markets for example), or lose their dominance of the hashpower.