The O(NlogN) algorithms all have very large constant factors - IIRC the crossover point from karatsuba or tooms to FFT based solutions (the NlogN algorithms) is in the thousands of digits.
The recent O(n log n) algorithm is only useful for numbers of size larger than 2^(2^(9^12)). Which is to say, this will never be of practical use in our universe. (Of course, conceivably someone else could invent a different algorithm of the same asymptotic complexity with a more reasonable constant factor.)
But it’s still nice to have a proof that this lower bound is attainable in theory. Makes it easier to write down the theoretical lower bound for a wide variety of other algorithms.
