To preserve the dynamics contrast accurately between pieces recorded at different times, I'd still advocate proper gain staging in all phases of creating the recording. Having all equipment and software setting fixed takes guess work out and ensures consistency.
Take the Zoom H4n Pro for example, you still want the signal out peaks at -6dB. H4n has a "REC LEVEL AUTO" feature that can help finding the proper mic level to ensure this. Enable it, put the recorder in recording standby, then play the loudest you think you'd ever play. The recorder will detect the loudest sound then set the mic level automatically so that the loudest sound will record at -6dB. Note the displayed mic level numeric value on the screen, turn off "REC LEVEL AUTO", then manually set the mic level to this fixed numeric value. As long as you don't change your recorder's position in subsequent recordings, keep this mic level unchanged.
Later in post processing, you can boost the audio level if you want the final recording to sound louder. If you did the step above correctly, you have 6dB of headroom to play with. Let's say you decide to make the peak of the final recording at -1dB - you simply increase the level +5dB from the original output of the recorder. However, the key is, you fix this +5dB the same for all of your recordings, not just the quiet ones. The goal is quiet playing should till sound quiet and loud playing still sound loud, and the contrast remain the same even between different pieces.
Now, with 32-bit float, what leeway you have is you can set the mic level higher, and reduce the level in post processing. One main reason to do this is to mitigate high noise floor in your recorder (a common problem for nature recording but less so for piano). Still, you want to keep mic level figured out, set and fixed for consistency.
The final question is what if you do all this, and the quiet playing part of the recording just sound too quiet when playback on a mobile device or in a noisy environment. Then you will need some level of compression to bring the quiet part louder. It's a personal decision, depend a lot on what your target audience and their listening environment is. Once decided, it's advisable to keep that part of your post processing workflow fixed as well.