My very humble point of view on 51% attack is BULLSHIT! unless someone proves it.
I am pleased that there are others prepared to call "foul" on this one.
I can't find a single scholarly article to back it up. All references lead back to a single source, often quoted verbatim. I remain a skeptic. I require proof.
One has to ask why it has never been demonstrated. Could that be because it is practically impossible? As I understand it, an attacker needs to 1) solve a minimum of two consecutive transaction blocks before anybody else does AND 2) convince the rest of the network to cooperate. Logically, it falls down at stage 1 because no matter how much hashpower you have, you cannot guarantee that someone else will not solve the block before you.
Feathercoin was assaulted with a 51-percent attack in 2013
http://www.coindesk.com/feathercoin-hit-by-massive-attack/