A few comments -
Audacity shows potential clipping. For example, if you have a non-clipped file and you amplify (and allow clipping) and you can make Audacity “show red”. But since Audacity uses floating point numbers for sample storage, there is virtually no upper limit. The peaks can go way-way over 0dB without clipping so you are just seeing potential clipping. If you play it at “full digital volume” your digital-to-analog converter WILL clip, and if you export to a “normal” (integer) WAV file or make an audio CD, it WILL be clipped. If you export as 32-bit floating-point WAV the file won’t be clipped, but again, if it’s played-back at full-digital-volume the DAC will clip.
Conversely, if you have a clipped file and you reduce the levels in Audacity, it will no longer “show red”, but of course the waves are still clipped.
MP3 can go over 0dB without clipping (but again the DAC can be clipped during playback). And, since MP3 is lossy compression it changes the wave shape… Some peaks get higher and some lower, and if you make an MP3 from a 0dB normalized WAV (or CD) the new higher peaks often go to +1dB or so (without clipping) and again you’ll “see red” in Audacity. The DAC will clip at full-digital-volume but as far as I know this clipping is not audible… If you hear a compression artifact you’re probably hearing something else. But, some people do like to leave a little extra headroom when making MP3s to prevent this.
This same wave-shape “distortion” can tend to hide clipping. The tops & bottoms of the clipped waves may not be perfectly-flat after MP3 compression, but that doesn’t fix the sound of the original clipping distortion. The process of “cutting” and playing-back vinyl records also changes the wave shape which tends to hide clipping (without fixing the sound of the distortion). With vinyl, the changes in wave shape can also gives you a higher crest factor (peak-to-average ratio) which fools some people into thinking the vinyl recording is less compressed. The sound of the dynamics isn’t changed but the measurement “improves”.
Most of the streaming services are now using loudness matching. That brings down the volume of most music so clipped music won’t hit 0dB and that will tend to “hide” any clipping. And again, the lossy compression will change the wave shape so you may not see squared-off waves in Audacity.
I tried the Audacity Clip Fix effect once or twice on a “Loudness War” CD rip. It made the waveform look better but it didn’t improve the sound quality. I haven’t tried it on a “cleanly clipped” recording… The results should be better.
It’s theoretically impossible to reverse clipping because the original height & shape of the waveform cannot be known. In some cases you should be able to get some improvement.
Somewhat ironically, some people have used more limiting, or a tape-saturation plug-in, to “round over” the clipped peaks and remove some of the distortion. But of course, that further squashes the dynamics.
Music mixing & mastering is complex and a “loudness war” recording isn’t JUST clipped. It’s limited & compressed. Limiting suffers from the same issue as clipping… It’s impossible to know the original height & shape. Theoretically, compression can be reversed* but in practice you don’t know the compression settings (attack, release, ratio, threshold, etc.). Multiband compression adds more unknowns. It’s also often an iterative with compression/limiting, EQ to “fix” some of the changes caused by compression/limiting, and then a little more limiting. That can’t be un-done. And, any compression/limiting done on the individual tracks before mixing can’t be un-done.
The bottom line is - You can’t really un-do the damage done to a loudness war production. Of course it’s intentional, the producers who make loudness war recordings and the consumers who buy them (and sometimes push them to the top of the charts) don’t see it as “damage”. 
\
- There are noise reduction techniques that use complementary compression/expansion.
