Miners can not operate anonymously nor do anything useful with selfish mining or double spends as long as they are mining on pools, which the huge majority of the hash rate is currently doing.
Incorrect. Just by being on a large pool, they inherently are mining on the correct block more often than the minority of the hashrate, which instead have to wait for propagation delay.
That's not evidence of Bitcoin operating outside a Nash equilibrium, which was CfB's argument. Most of the hash rate 80%+ obviously mines on very large pools. The rest may do something else, and it might even be a bad decision on their part, but that little 20% doesn't really affect how Bitcoin operates.
Also a mining farm mining (or oligarchy of them) with 33% of the hashrate even if mining on a pool, can still do the selfish mining attack without informing the pool. Remember Bitcoin never implemented Meni Rosenfeld's oblivious mining shares fix to prevent share withholding attacks. This is assuming the pool supports
getblocktemplate so the miner can set his own transaction block hash.
I don't know which pools support getblocktemplate, but there's no evidence of anyone using it (most if not all ASIC gear just connects directly to the pool and uses stratum). Seems like that concept mostly died when Luke-Jr lost the original argument with slush.