So long as:
- The Gimp Import format matches the Gimp Export format,
- The image format is not a compressed format,
- The Audacity Export format matches the Audacity Import format,
- The format in Audacity is mono and not compressed,
you will usually get ‘something’ that is at least vaguely meaningful.
The way to test is to export from Gimp and write down the export settings, then import (RAW and mono) into Audacity and write down the import settings. Then, without doing anything to the audio track, export from Audacity using the same settings as the import, then import (“open”) in Gimp using the same settings as the export, and the image should be roughly the same. The difference between the new image and the original image shows how much data loss there has been.
For best results, turn off “High quality conversion > dither” (see: Quality Preferences - Audacity Manual)
Headerless formats are generally better because then you are only dealing with the image data.
You will get different results when processing depending on the formats used. There is no direct correlation between pixel values in the image and amplitude or frequency in the audio track, it’s just a relationship between pixel values and sample values, so the process is all rather hit and miss.