Both methods = applying gain and automation get you to the same place and each has merits.
As said the quick and easy use of a volume envelope is simple and non destructive. You don't like the results, no problem drag the node to fix.
But then volume envelopes have a down side when you start chopping songs up and moving things around.
Often the reason a vocal part is at a different level was a mistake made will tracking. Say a punch in and the singer moved further away from the mike.
If it's only a short section you'll often actually "see" the difference in level when looking at the waveform.
As the OP said, they have back ups so there is no real reason not to perform a gain rendering to that little section.
You dial in the the amount needed to level that part with the rest and render and forget about it. It's done. No messy envelopes. And the word destructive is misleading. We are rendering at 32 bits.. I dare anyone to find anything that's been destroyed.
Now you slap an compressor in the bin and push the vocals and all will be nice and even.