Nakamoto coefficient is way higher than 2

People who say so keep considering pools as a unique entity. While it is true that someone leads the pool, it is wrong to consider all miners to blindly follow regardless of what happens.

Let's consider an attack on the Bitcoin Network of more than 3 blocks reverted :

If a pool leader was to attack, following miners would mine blocks on a hacked blockchain. After a few blocks, the hacked blockchain might become the longest one (difficulty speaking). The hacked chain would get considered as a fork by all honest nodes because it would imply a reorg higher than 3. Honest nodes would automatically discard the hacked chain. The word would spread on all social medias than an attack is going on. Everyone would manually check they are on the good chain. At the same time, miners in the hacking pool would start taking action. No miner would want to mine on a worthless chain, and a majority of miners are honest. Thus, they would quit the pool. The power of the attacking pool would fall dramatically. After that, the fork would not receive enough difficulty mined to keep up with the good chain, and would die.

I think this whole attack cancellation could be put to work in less than a few hours. Plus, the attackers would lose insane amounts of power to mine on a chain that loses in the end.

Let's remember than the majority of our miners want to make good. If they didn't, we would punish them anyway. So they will never tolerate their mining power to be used by a malicious pool. As soon as a pool is labeled as malicious, it's over. That makes their leader completely powerless, and that makes proof of work the system with the highest Nakamoto coefficient achievable.