Smooth, you've constructed an elaborate strawman.
I claim the entire purpose of crypto-currency is a) permission-less commerce and b) decentralized control to prevent gaming the control over the issuance of money (so it can scale globally among other benefits).
Since I already argued that one can't mine with lower economies-of-scale without losing hash rate share over time, then the ability to mine or not mine in any specific pool is irrelevant to homeostasis of the case #b.
So that leaves us only with #a remaining as it pertains to pools.
QED.
For his original argument to hold he has to show both that the distribution-of-power outcome is different between delegates in DPoS and pools in "delegatable PoW" and that the DPoS outcome is (in some specified way) preferable.
It consumes less electricity.
Fact is that mining will become ever more centralized in Satoshi's design because of the economies-of-scale of ASICs and electrical power. I believe there was maybe even a research paper that proved something along these lines?
The fundamental problem is the mining is done for profit. For as long as that is the case, ASIC farms (or Larry Summers' 21 Inc. economies-of-scale) and subsidized, industrial/government/utility scale electricity will rule.
Also due to bandwidth issues and that every full mining node has to validate every transaction, scaling transaction volume will force centralization.
Also your argument about sacrificing cost is nonsense, because the low cost leader will take hash rate from the others over time by reinvesting higher rates of return.
Edit: Smooth has a valid point if no entity (or collusion of entities) has 51% of the hash rate because then someone could sacrifice mining losses in return for censorship resistant way to post transactions to the block chain. So in that sense, my word "nonsense" is incorrect and I apologize. But the huge glaring flaw is that once the State can regulate 51% of the mining power (which is destined to be centralized), then Smooth's caveat no longer applies. And this is my overriding concern, so that is why I often downplay this caveat that smooth points out.
Edit#2: however if hashrate is very large then smooth's caveat is really pointless because who has enough hash rate to push their transaction onto to the block chain without a pool. And again Satoshi's design doesn't enforce that pools must allow getblocktemplate. If your hashrate is not too small, you can just mine on any pool that offers getblocktemplate and wait a long time until you win a block solution to insert your transaction, or just mine a long time solo. Many could potentially join together to pool their resources to mine at a loss to have ready access to censorship resistance, but unless you are using P2Pool (which can be attacked with share withholding attacks) then the State might target your pool server (but again I think it is easier for them to just target 51% of the hash rate for regulation requiring all transactions to carry KYC, since you might place your server behind an anonymity network although this will be very difficult to do in Satoshi's design because of the bandwidth requirements).