I know you’ve decided to kill this off, but did you ever look into getting raw xml data from the codex and using xsl to transform it into a pdf file? It may take a little bit of work, but you could write a script that compliles a new help document at given intervals.
There may be other factors I’m missing here.