I would choose bounce over flatten for all aspects of audio pitch editing, because it is easier to cut and paste a small snippet of audio than it is to delve into edit history.
My usual workflow is:
1 Make copy of untouched vocal performance and put aside as safety/ restore option
2 Bounce sections of audio, say 30 seconds roughly (long sections make for slow graphic response in variaudio)
3 Work on variaudio, pitch, stretch etc until satisfied
In most cases I leave things at this stage. Sometimes I might duplicate the edited track and once again bounce audio.
This gives me the choice of being able revert, compare and rework small sections. With everything on screen, labelled and clear.
P