Of course, but what happens when they selectively raise the renewal fee for your domain to $500 quadrillion? (Also, what prevents them from changing the controller back, if they have 51%?)

The renewal vector is the only way atm, yes. But it's easy to prevent this as a .eth holder: just register for a really long time. Costs like $500 to register for 100 years for most names.

As for the controller bit, even if you change the controller, we can't revoke the .eth names. We can only change new registration parameters (as you mention).

i haven’t kept up with ENS to this deal. they moved from a multisig to a DAO. as GP claims, they also restricted the set of actions available to governance before this move. on the face of it then, it doesn’t look unlikely that the DAO will vote to further restrict its own abilities as time goes on. this exact mechanism has been used to progressively decentralize other protocols, and as we see with other “governance tokens” (e.g. Uniswap — where IIRC governance can tweak like one single boolean parameter), massively restricting governance abilities doesn’t destroy the viability of the token (which is to say, it’s not clear to me in which direction the finance incentive — if any — points with regards to such “ungovernance”).

^ this, except I'd add that as a large ENS delegate myself (as one of the founders of Metaphor), I and other large delegates don't factor in the price of the token into decisions.

We never intend to sell $ENS, it's a public good, not a profit-generating protocol. Think of it like an on-chain B-corp.

That said, if you really want to make sure you can't ever face the "egregious renewal fee" attack vector, you could just register it for a really long time. For most names, it costs like $500 to register it for 100 years.

This wouldn't happen as it's against the economic interests of the ENS DAO. Similar to a company maximizing the value of its stock, the DAO aims to maximize the value of the $ENS token by being a good steward of the project.

The whole point of a DAO like this is to try to create good incentives for the project without central control.

Is that always true, or just in some cases?

For example, a website hosting child pornography I think would be quite detrimental to the value of the project. Would the ENS holders seek to maximize their economic value, or "property rights"? Or is child porn good for business?