I've seen a number of comments here claiming that, if you want an mp3 file to have a gain value not available in increments of 1.5 dB, that it's better to use the scale command during encoding (e.g., --preset medium --scale 0.83) than to use WaveGain to lower the volume of the wav files before encoding them. The justification for using --scale instead of WaveGain is that WaveGain is a slightly lossy process, while --scale is not.
However, it seems to me that the processes would be equivalent. That is, WaveGain calculates new sample values that are some multiplicative of the old. Which is exactly what I expect --scale does to the sample values as it is encoding them to mp3. The only way I can see that --scale is preferable to WaveGain is that --scale might not be limited by 16-bit audio, while WaveGain necessarily transforms a 16-bit wav into another 16-bit wav.
Any illumination or debate on this point would be helpful to me and others. Thanks.
[And yes, I realize that the new ReplayGain APE tags allow infininite incrementalization; I would use them if there were an easier way to modify them, so that I could set a louder album's target bitrate to 90.5 dB, for example. If there is an easier way to do this, I'd appreciate the knowledge. But even so, most current mp3 players won't read APE replaygain tags, so I like having a file that ahs the volume I desire independent of tags.]