Internally for quality reasons, I've always used PCM/WAV, for many of the reasons you mentioned. Its much faster to edit, and you do not have to worry about generational losses and artifacting that occur with compressed formats. That being said, when I record with one of the WinRadio's, it will create a wave file, I'll then go and edit the parts I want to save, and then archive it in MP3, usually at 128 mono. It will then play on just about every device at my disposal, even the cars.
On a similar note, about 70 percent of the current X-FM music library is PCM. When source material exists, the rest will be converted to PCM too. Its much easier to process clean source material than lossy material, and its one of the reasons the station sounds as good as it does. I also archive all my board feed recordings in WAV as well.
In regards to codecs screwing up weak signal details, that's what codecs do. The algorithm in the codec, through a process known as "perceptual coding" has to determine if parts of the spectrum are audible or not. In the presence of noise, the codec begins "guessing" what's perceptible and what's not. That's one of the reasons cymbal crashes and white noise sounds so strange on MP3 and other lossy codecs. Only so much bandwidth is available, and the codec has to figure out what to keep, and what to throw out.
+-RH