To attack the first question (dumping AUP to WAV), you will get either a small degree of loss or no loss, depending on your source material and the bit depth of the WAV file.
In general audacity defaults to 32-bit floating point format for audio data, where as default WAV export is to 16-bit PCM (integer). So there is a reduction in dynamic range in going from an audacity project out to a 16-bit WAV file. You would be relative hard put to notice this, unless you had recorded well below maximum level and then amplified the signal after re-importing the WAV file. This would raise the noise floor in either case, and so make the difference between the 16-bit recording and 32-bit float recording more obvious.
If the source for the audacity project was 16-bit however, and you haven’t applied any effects to the recording, then there would be no loss in exporting a 16-bit WAV file, because the extra precision of the 32-bit project is only carrying noise. So in that case there would be no loss at all in the 32->16 conversion, because it comes after a 32->16 conversion as the audio comes off the sound card.
The other way to guarantee that there is no quality loss when going from an audacity project out to a WAV file is to export a 32-bit float WAV file. The exported data then matches the project data format, and there is no quality loss at all. This applies regardless of the source of the audio and any effects applied.
I don’t know why you seem to have a 6:1 ratio between project data and exported file, unless you are looking at the project whilst audacity still has it open (assuming the project doesn’t have multiple tracks). If the project is still open in audacity then the full undo history will still be present in the _data folder, and so make it much bigger. So if you already have twice as much data because of 32-bit recording, then it only takes one duplicate of the whole project (from an amplify or noise removal effect) to get most of the way to your quoted size. This will all be deleted when audacity closes the project, or you manually flush the undo history (View > History dialogue).
Tackling the second question about 16-bit WAV to audio CD and back again, the answer is again maybe, maybe not. Audio CDs are 16-bit, so in principal the burn and rip process doesn’t loose any data. The problem is that audio CDs were not designed with archival use in mind (they are meant as consumer audio media, remember), and don’t have data-grade error correction (hence why you get more audio on an audio CD than you do WAV files on a data CD). So it’s rather more likely that you will get a disc burnt with minor errors that the CD player will cover up, but not actually be able to correct. The same applies in the rip phase, especially with CD-ROM drives designed for data use (where the disc has seek marks in the data), which often have a hard time extracting audio (with no such seek marks) and can easily drop or duplicate audio samples with no easy way to check. A good extraction application like CD paranoia will compensate for this by reading the disc multiple times and comparing the results to get the accurate data as far as possible, but it’s not foolproof.
So most of the time you will get perfectly good results from burning a CD and ripping it, but I doubt that you would find the results were bit-accurate.