I guess the only thing I want to reply to is point 2 about centralization. Yes, megagigaterapetablocks will likely result in fewer fully-validating, non-mining entities. It will also likely result in fewer full-stack mining entities. So what?
People keep talking about decentralization as if it is an end in itself. Why? AFAIC, as long as there are no structural barriers to entry by new participants, the network is as decentralized as it need be. If there is no discernible marginal benefit from adding one more participant to the network, then by definition further decentralization is of no benefit.
And quite frankly, all public blockchains -- to the extent that said blockchain is economically significant -- will centralize thusly - at least with respect to mining. How many parties make up 51% of BTC hashpower? Four. Already.
Not to mention the fact that non-mining nodes provide zero benefit to the network anyhow. If you have sufficient economic interest to be a first-class tx creator, then fine. Pony up for a validating client. But don't delude yourself that your participation brings any benefit to anyone but yourself.
By having megagigaterapetablocks you are giving full control of blockchain audit to the "operators". I guess the only way some people would understand why centralization is bad is when forces collude to censor certain transactions, such as donating to Wikileaks or anything else that wouldn't comply with what they consider a morally correct transaction.
All public blockchains -- to the extent that said blockchain is economically significant -- will centralize thusly - at least with respect to mining. How many parties make up 51% of BTC hashpower? Four. Already.