Fact: if the entire consensus mechanism as it was designed, is decided by 20 entities, it is not a peer-to-peer CONSENSUS SYSTEM any more. And that it the current reality.
Which, sadly, implies it's a failure. "The system is secure as long as honest nodes collectively control more CPU power than any cooperating group of attacker nodes.". Course one could "argue" about what the definition of "attacker" is but it is what it is at this point.
I think it is quite clear what an attacker is in bitcoin's case: it is an entity wanting to modify the published consensus history.
In other words, a miner building on top of the chain is never an attacker (a miner orphaning the last block isn't, either). Someone trying to overdo the last 50 blocks, however, is an attacker.
I think one could define an attacker as someone orphaning ON PURPOSE any block older than 6 confirmations.
It is here that proof of work is a terribly BAD cryptographic security. You can't find worse. In as much as consensus FINDING can be done with just any sufficiently fair and random way, past consensus securing should be done with better cryptographic means than proof of work.
An attacker that controls more than 50% of the network's computing power can, for the time that he is in control, exclude and modify the ordering of transactions. This allows him to:
Reverse transactions that he sends while he's in control. This has the potential to double-spend transactions that previously had already been seen in the block chain.
Prevent some or all transactions from gaining any confirmations
Prevent some or all other miners from mining any valid blocks
I do agree, given what we now know for a certainty, that the bitcoin method of using PoW for "all things" is a major flaw.
In fact, as long as these 20 entities are not attackers, and they have never been
As of today. But that "20" will continue to shrink until someone can control 51%. It's inevitable assuming "all things" stay the same.