I found this scholarly article to be a concise summary of many interesting points: http://www.isi.edu/natural-language/people/voynich-11.pdf ("What we know about the Voynich manuscript" by Reddy S. and Knight K.).
Rugg (2004) claims that words might have been generated using a ‘Cardan Grille’ – a way to deterministically generate words from a table of morphemes.
(Gordon Rugg. 2004. The mystery of the Voynich Manuscript. Scientific American Magazine.)
If you know there was a reasonably fast way to produce the content which has exactly observed properties, why searching for some more complicated explanation?
But why generate all that text using a cardan grille?