Since Audacity has its own web page (on Sourceforge, if memory serves), and they have discussions there, you could contact their developer and suggest anything you like there. That application is not specific to Librivox, nor is Librivox limited to the use of Audacity for recording, so we make no particular effort to collect suggestions on it, methinks.
(thanks to Availle for the links!)
That said, I don't think that a library of sounds is going to help if in one's recording one misses some phonemes.
The problem is not the [in]ability to record (or extract) those phonemes and store them in a "library", the problem would be to paste them back into the track in such a way that they don't sound unnatural. Have you actually tried it? When something is missing or wrong (an article or a preposition) in my recording, for instance, I usually end up re-recording at least the entire sentence or phrase in which the error occurs. It's because the transitions between phonemes is often impossible to fake by editing. Perhaps I'm not good enough at my editing, of course; it seems easier to me to re-record than to spend the time figuring out how to paste a single phoneme and then spend the time every time when actually doing it...