I am not too familiar with MP4 (and the sound encode there), but for an mp3, no re-encode is necessary to normalize the volume as it is set in the header. What puzzles me though is the +10 db whatever used here. I am more used to normalize in percentages of maximum 100% (0)
I do normalize custom collections of music I put together and so it is mostly peak maximizing to about 95% (this is to encompass processing on the fly which can add another 2-3 db gain and so to avoid clipping).
Since most of it is flac, they are temporarily converted to .wav-normalized-reencoded to .flac and so no header-normalizing as would be the case for MP3. As is evident, I am a tad old school and make all collections physical collections by copying the actual tracks instead of just creating playlists as it is easier to transport collections for reuse that way.
This was why I said I didn't quite understand the question as a +10db gain would cause clipping if the file already peaks at 95%.
And just for completeness, Album-gain or average I would use for old albums to keep the level between songs. A typical scenario would be a non-remaster cheap production of a '70s recording with too low volume.