This is a simple, and thus great, way of using Makers incentive structure against itself.
While specific to the MakerDAO ecosystem, it effectively highlights the complexity and challenge in adequately creating a decentralized governance structure.
Essentially, to remove centralized control, you must structure incentives such that the system ultimately acts as intended, but there is always going to be ways to subvert the incentive structure. The goal is that costs of conducting such activity are too much to be reasonably conducted.
It seems that these systems will take years and years in the wild before their incentive structures are iterated upon enough to finally get to a point where truly large business value will flow through it. Exercises like this are useful in either disproving this whole concept or incrementally improving it.
You basically described PoW (proof-of-work) mining, which was (imho) the actual innovation with Bitcoin. The current idea though is to build a byzantine fault tolerant system that doesn't require massive amounts of "wasted" power. I put wasted in quotes because it is actually doing something, it is just the output isn't useful for much more than validating the transactions.
Interestingly, Maker is currently built on top of (Ethereum) PoW. Some day, ETH2, which is PoS (proof-of-stake), will solve all the same issues you're talking about. Or at least that is the dream. We will then have a pretty amazing digital rube goldberg machine where we can sit back and watch the gears turn.
PoS will not solve anything. PoS is just obfuscated proof of work. The same amount of "resources" will be wasted in a PoS system, its just going to be a different type of resource.
I'm not in it for the audience.. in any case, maximalism is mostly a self-assigned title and the author of that blog describes himself as such
edit: how can something that is being discovered be late? The work on PoS is almost completely greenfield and far from trivial. Many people made many far too optimistic comments about how long it would take to get it right but no one can say it's late...
Even with HBM on the card... while it will hash fast, you're still going through the memory controller. What you need to do is skip that. Get rid of the memory hardness entirely. Generate the DAG on demand.
The analogy with the Prisoners' Dilemma is interesting here, but doesn't feel quite right. In the comments, BenWang summarizes it like: "the effects of the attacker will happen to all token holders regardless of what you decide to do with your tokens, therefore, you might as well rent them out."
It seems that the MKR holder in this case could rent them out... or sell MKR and exit their position entirely. And this seriously discounts the fact that eroding Maker security is immediately obvious and can get priced into the MKR token very quickly, defeating the point of a few percentage points of yield.
It makes sense on some conceptual level, but it's not clear that the magnitudes of the incentives result in this toxic game theoretic equilibrium.
While specific to the MakerDAO ecosystem, it effectively highlights the complexity and challenge in adequately creating a decentralized governance structure.
Essentially, to remove centralized control, you must structure incentives such that the system ultimately acts as intended, but there is always going to be ways to subvert the incentive structure. The goal is that costs of conducting such activity are too much to be reasonably conducted.
It seems that these systems will take years and years in the wild before their incentive structures are iterated upon enough to finally get to a point where truly large business value will flow through it. Exercises like this are useful in either disproving this whole concept or incrementally improving it.