I will wait until 2018 for the next Fields award instead. Seems to be cheaper. To me it is like those headlines nowadays: you get hooked, and if you have a closer look, it results in bare disappointment and anger about the wasted time. I can't imagine such a result in a textbook without any earthquakes far ahead of it. Even Wiles created tsunamis although his proof was understood by at most a dozen people at the time. (Not sure whether this has significantly changed.)
If I remember correctly, then NP can be done in polynomial time if one allows additional means like oracles or something. My bet would be, that the author(s)' arguments go along with such extensions, e.g. quantum computing or restrictions to incomplete NP problems. There has been a theorem on graph isomorphisms recently which pointed in a similar direction, of course without solving NP = P.