A lot of this talk of decentralization is also practically meaningless unless that element of the network is put to the test. Bitcoin could operate on 1 mining pool, and would this prevent its price from being above $1,000? Would it prevent sending a transaction?
I understand the point here that decentralization reduces risk and provides greater security, yet it is all talk if it never matters for the practical use of a given system. If ETH was fully centralized, you could still run your smart contract on it. What you could never do on it, is have 3 second response times and 10,000 TPS recorded on the blockchain. EOS is more decentralized than these alternatives, and is decentralized enough. But whatever the case, neither Ethereum, Bitcoin, or EOS are likely to face the full-on stress attack against the system, where someone spends millions/billions to try to subvert these systems.
If you wanted to do this, for BTC & ETH, you would target the mining operators and the nodes. These are large companies and businesses, and very likely you could have it completely controlled by influencing less than 50 people. Certainly the network could be attacked by just going after the few mining pools to achieve more than 51%.
Vitalik going after "decentralization" is really a bait and switch against the weaknesses of ETH which is a huge system bottleneck, and too slow for a lot of practical uses.