a consensus needs 95% to switch
No. Kiklo spouts a lot of nonsense, presenting merely possible conclusions as matters of settled fact. But he or she is indeed correct that any group in collusion holding greater than half of the mining power can rewrite the blockchain*.
The '95% consensus' is merely a self-imposed threshold.
*Of course, game theory points out that they stand to lose their vast investment in hashing machines, as users abandon a broken system.
Further, they still have not clarified what the heck they are spouting about by saying "it was the Chinese mining pools that blocked the proposed updates". But again, I've kind of lost interest in that particular discussion.