How could you prove that VM is gibberish?

[Koen G]

I think there is no practical way of doing this. This is why it's so easy to blanket dismiss all meaningless hoax theories as "mere conjectures". Unless you find an undeniably original signed affidavit from the author of the manuscript describing its specific characteristics in great detail and providing a reasonable account of why the meaningless manuscript was created, I don't think there ever will be a sufficiently strong argument for any meaningless MS theory. Even if the text was generated using some stochastic approach (and we reverse engineer this approach), all it shows is that a manuscript like this can be created using this approach, not that this manuscript was definitely created using this approach.

I agree, and there's a good chance we'll be stuck in this limbo forever. There are many scenarios where the MS is meaningful but we are no longer able to recover that meaning; basically any scenario where additional, now lost information is required (like a codebook) or in the case of lossy/one-way ciphers. And I don't think we can ever distinguish the output of such scenarios from a hypothetical (stochastic) way of generating nonsense text.

Basically, the only case when we might ever be able to tell whether the MS is gibberish or not, is when it is meaningful and that meaning can still be recovered.

I also think what Mark proposes is attractive in some way: even if the MS is meaningless, there may still be actual words or sentences hidden in there. But this would only complicate matters, because it would mean that the creators of this system not only came up with a completely unique way of generating large amounts of cohesive gibberish governed by all kinds of rules, but they also snuck in a way to use the same system for enciphering actual text. So even though I like this solution on an intuitive level, I think it is not the most likely in practice.