Rats - I posted an extensive reply, but it’s disappeared 
Bug 87 is certainly related, but there is a common “special case” that can be addressed in a different way. This “special case” is when normalizing “s”. This is not a cure-all solution, but would be a useful feature. In some situations, if the peak value of the selected audio is known then the peak value of the output can be calculated (either exactly, or as a ball-park figure, depending on what processing is being done). The peak value of “s” is known to Audacity, so why not make that value available to Nyquist?
More about this here: Normalizing
Replacing “s-read” so that Nyquist can read audio data from Audacity (as mentioned by Richard) would address the performance limitation, but I think that passing the peak level of the audio to Nyquist is more of a feature request.
I’ll get back re. bass/treble effect in a new post.