It’s not just the quality of the vocals in itself but also the vocal delivery. A good example is the demo for For All Time compared to the finished version. The demo has different phrasings, different pronounciations and a slightly different melodic rhythm on certain words that did not flow as well as the changes he made for the final version for For All Time.
The same can be said for the toy/ best of joy. His phrasing and rhythm flows much better in best of joy than it does on the toy imo, so it is understandable that he wanted to work on those elements. So instead of just punching in a single line here and there he opted for re-recording the entire thing which, in most cases, makes it less jarring.