* It's possible - very likely even - that even if somehow P=NP, the fastest algorithm for any NP problem turns out to be something like n^1000, which is technically P, but not practical in any way.
* The proof may not be constructive, so we may just know that P=NP but it won't help us actually create an algorithm in P (nitpick: technically if P=NP there's a construction to create an algorithm that solves any NP problem in P time, but it's extremely slow - for example it involves iterating over all possible programs).
* It's possible - very likely even - that even if somehow P=NP, the fastest algorithm for any NP problem turns out to be something like n^1000, which is technically P, but not practical in any way.
* The proof may not be constructive, so we may just know that P=NP but it won't help us actually create an algorithm in P (nitpick: technically if P=NP there's a construction to create an algorithm that solves any NP problem in P time, but it's extremely slow - for example it involves iterating over all possible programs).