Consider the analogy of a parking garage in a very convenient location (say right next to a popular theater) when there is free parking available a short to moderate distance away compared to the situation with the same garage but no free parking available at all. In the first case, the garage may charge only a nominal fee and nearly everyone (or conceivably everyone) might pay it for more convenient access to the theater. In the second case, the garage will charge the maximum fee possible until people stop going to the theater at all.
Except as your analogy applies to Satoshi's proof-of-work design, then the free parking is not accessible by anyone who has a car because it is on the top of a skyscraper[1]. The only people who can access this free parking must either have a helicopter or they must pool their resources to buy one.
You are confusing miners and users. Perhaps they could be the same, but they need not be. (Indeed satoshi's original design more explicitly divides the two than current thinking among many prominent Bitcoin developers.)
r0ach's claim was that pooling of mining
in practice makes the consensus system of Bitcoin inherently
delegated, but that is false because miners need not pool, and certainly need not pool with one of a fixed set or even a fixed-size set of pools. Thus this is
different in structure from a system where delegation is required. (It is true if you posit that Bitcoin
will certainly be 51% attacked, as I think you believe but can't prove, and turned into a centrally-controlled system instead, where pools or whatever they are they are called in that outcome enforce membership, but then why bother with the delegation argument, just show that this is certainly true and prove Bitcoin useless unconditionally.)
I specifically doubt that the outcomes will be the same in a system with
permitted delegation to an open set of delegates (pools) as opposed to a system with
required delegation to a fixed size set of delegates. But if r0ach wants to try to prove the same outcome in his war game model, he is welcome to do so, and I will read his proof with interest. Unfortunately if he does
that, then he will show no advantage to DPoS over PoW, which undermines his original argument.
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.