^^ the thing about 51% attack is that it can only be considered an attack and also be profitable if it is accompanied by an actual double spend otherwise just replacing an already mined block with a block the attacker mines has no benefits with the same exact transactions has no benefits.
when speaking of profit you first have to factor in the cost. and since the cost of 51% attack is in the millions of dollars the transaction that is being double spent must be worth millions of dollars.
now that we establishes the tx value we have to think of a trade that involves a transaction worth millions of dollars and the other party doesn't ask for a high number of confirmation such as at least 30. such trade does not exist in real world. nobody has ever made such a transaction and only accepted low number of confirmations ever.
with that kind of high confirmation requirement the cost of performing a successful 51% just shot up into billions of dollars instead.
IMO if someone attempt to make double-spend attack, most likely they would try to fool multiple parties and using multiple identity (e.g. deposit smaller amount of Bitcoin with multiple account on exchange rather than make a huge Bitcoin deposit to smooth the attack).
That "economically irrational attacker" would be wasting resources. The network is made of people, capable of coming to consensus if there was a direct threat.
What the troll said was that the miners could collude. But would they? They would be opening a strong debate for a POW change.
How about scenario mentioned by @BrewMaster?
I already said it, that the Bitcoin network is composed of people. They would be wasting resources. POW change to hard fork the network away from bad acting miners would be strongly debated.