Here's a link to Elias Schwerdtfeger's very interesting "Das Voynich Blog".
Elias has worked really hard behind the scenes to find ways of visualising the statistics expressing what "old hand" Voynichologists (such as, say, Philip Neal & I) see when we look at the Voynich - you know, the highly bonded, multi-level internal structure that exists at the stroke, character, glyph, digraph, word, line, paragraph, page and section levels.
As an aside, I've long disagreed with Renaissance encipherment hypotheses for the VMs based on moving alphabets, specifically because they fundamentally destroy these kinds of internal structure: the only way to keep such hypotheses alive is then to argue (as, for example, my old friend GC does) that these structures are part of the "surface language", i.e. that the encipherer is dynamically stretching his plaintext to mimic these structures in the ciphertext. Yes, it's possible, but... put all the pieces together and it's a bit too much of a stretch for me.
Incidentally, I've been looking at f2v recently, specifically because of the "fa" marginalia there (one of the very few marginalia I didn't really cover in my book, "The Curse of the Voynich"). Elias discusses f2v at some length, proposing the eminently sensible (and testable) hypothesis that the same pe(rso)n that/who made the dubious (o)ish(i) emendation to the last line of f2v also added the "fa" marking above the second paragraph. They're both in similar darker ink (which is a good start): but I think that the Beinecke's scans - though fantastic for most purposes - fall just short of being able to resolve this kind of question definitively.
Actually, I've got a list of about 50 similar/related cross-indexing questions like that I'd like to address (say, by multispectral imaging or Raman imaging) in the future. But for now, that project is stalled (because the Beinecke turned my proposals down). Oh well: maybe next year...