I would consider working on the processed stereo file first, with the processing on a stereo aux or master mix bus. Processing the M and S separately can be useful for fixing particular problems, but if overdone can make the stereo image shift around, so if the stereo processing approach isn't cutting it for all problems, then go back selectively to the mono files, that way it's minimal. Of course you may be essentially doing exactly what I describe. I'm not sure what tools are available in Audacity, the program I use has a mid side specific dynamic EQ that also allows for a HPF to be used on the side, so I am hearing it in stereo as I do it. If you aren't monitoring the mono changes in stereo, maybe consider setting up routing so you can at least do that.