Phonology of Voynichese

[MarcoP]

If the author repeated certain words in the text extremely often, to a degree that is not typical of most written texts that we are familiar with, and if those certain words that are repeated extremely frequently happen to end in a final "-s" in the underlying language of the text, then this feature of systematic exact repetition of words could indeed also cause and explain an unusually high proportion of the total number of words in the text that end in a final "-s".

Thank you for the clarification. I think the idea makes sense: it should also be possible to measure if -y words tend to repeat more often than other words. Exact repetition is too rare to have a significant impact, but maybe quasi-repetition could? Anyway, this doesn't tell us anything about the phonetic value of -y, so I am not sure it is relevant to the present thread.

Importantly, I also note that if certain conjunctions, prepositions, articles, and other particles that are typically written as separate words in standard Greek text, are rather in the Voynich MS typically written as prefixed elements of the following words, then this could dramatically reduce the frequency of vowel-final words in the whole text, since many of these words are vowel-final, which could in turn dramatically increase the frequency of "s"-final words, for example.

Again, it can be true that, for some languages, particles written as prefixes would significantly alter word-ending statistics. It could also be that in Greek this would lead to a higher frequency for -s, but that should be supported by actual numbers. Anyway we are again entering into subjects that are both very specific and very speculative and little related to a general approach to the phonology of Voynichese.

It is well known in linguistics that while vowel sounds are more important for recognizing words heard in oral speech, on the contrary consonant letters are more important for recognizing words that one reads in written text. This is why all-consonant abjads exist and have long existed historically, while nobody ever heard of a writing system that only represents the vowels and not the consonants. To see this, just compare "lng exstd hstrcll" and "o eie ioiay".

The point is, if a particular writing system has the problem of a small number of characters to represent the entire language, it is much more important to reserve as many of them as possible to represent as many consonants or groups of consonants as possible, rather than to use too many of them to represent vowels. We can afford to sacrifice the representation of a vowel or vowels much more readily than we can afford to sacrifice the representation of a consonant or consonants.

The idea of an abjad could well be one of the most important aspects when considering Voynichese phonology.
If the writing system is not an abjad, vowels and consonant should alternate; they should cumulative have something (very roughly) comparable with a 50-50 share of character occurrences (not types: vowels will have less types, with more occurrences per type). That is basically the principle underlying both Sukhotin's algorithm and Hudson's OCP software (and possibly 2-states HMM as well).
As I wrote above:

I believe that, if the script is phonetic and it is not an abjad, there is little doubt that EVA: e, o, a, y correspond to vowels

This opinion is based on accurate analyses by several researchers, using different transcriptions and different methods.

On the other hand, if the script is an abjad, it might be as you say. I don't know much about abjads, but if vowels can sometimes be totally absent, then (if Voynichese is such a case) y would obviously represent a consonant. 
Abjads are a subject I would like to investigate more. For what is worth, the fact that Sukhotin's algorithm results in a higher Consonant-Vowel alternation rate for the VMS than for the comparison languages I have checked (German, Italian, Latin, English) suggests that Voynichese does represent vowels.