I don’t know about Audacity, it’s ok for simple work, but it doesn’t perform so good when resampling, so keep in mind that every step you take destroys a little bit of your assets’ fidelity.
Here’s some comparisons of noise introduced in SRCs with Audacity and Reaper (which also features a great batch processor):
Here is a clearer image of the noise and alias frequencies that are generated by each program:
Again, Acon Acoustice against Reaper:
And the same programs in the noise spectral view:
The latest Adobe Audition also features awesome performance on conversions of such kind:
Of course you might not do any SRC during your workflow, but sometimes you can normalize and export at the same time, so usually game developers do that kind of mistake. And by the looks of it, it seems that Audacity, which is great as a free alternative, might have low quality mathematics/processing in other modules too. So it might be better to find an alternative if possible.
By the way, @Feuertopf1 do you want your files to be normalized by their peak values or perceived loudness?