My experience in working on 3 different pianos, and a choir library is that you get a better result if you NR on a per note basis. I believe this is because the notes are not necessarily recorded all in one session, and therefore there will be other external noise, slightly different noise signatures at play. This will be true even if they are all recorded in one session.
Of course, this will take longer, but I think the results are worth it. The other thing is, you might like to just try a small section using each method and decide for yourself. It might be the case that your own situation is different from mine and you get good results by processing the whole recording in one go.
Another thing I have found is that using the noise fingerprint track, or “wild track”, is also less accurate than using a small section of noise before each note starts to use as the noise fingerprint. Again, small variations in external noises, body movements, microphone movements (even small) will impact on the success of reducing the noise floor.
But, again, feel free to experiment as you may find you get different results.