I'm sorry if I'm saying what you already understand, but it definitely should improve the sound quality, simply because there will be more "meat on the bone", so-to-speak.

When you stretch a 44 or 48k file from 120 BPM to 100 BPM, for example, the samples have to spread out, and the software has to fill in the gaps. The more you stretch, the more it has to create. This causes artifacts.

Now when you have a 96k file, you have twice as many sampling points, so the software has half of the amount of space it needs to fill in, causing far fewer artifacts.

The same should apply to the pitch shifting as well: The frequency changes affect the distance between sampling points. More samples, less artifacts.