Syllabification

[Sam G]

This is a verbose cipher... every instance of the same plaintext word would yield the same sequence of ciphertext words

Not exactly - note that in my example "oolooooo" can be reduced to "oola" (as I did), but also to "alooooo". Which would be two different ciphertext words conveying the same underlay character (i.e., space). If we use this trick not only for "o" sequences, but also for "l" sequences, introducing a fourth letter into the alphabet, the degree of variability will rise further.

It would be apparent pretty quickly that "a" and sequences of "o" are equivalent and that would allow the repeated sequences of words to be uncovered.

Basically, what makes the ciphertext "unpronounceable" is the comparatively low number of letters signifying vowels in the alphabet as related to consonants. I think that if one uses an alphabet with roughly the same number of vowels and consonants, any ciphertext will be more or less "pronounceable".

I basically agree (though I think you also need at least some regular structure to prevent occasional long runs of consonants), but the VMS text is actually pronounceable with far more consonants than vowels, which is going to be harder to achieve.

Another cipher technique for making the unpronounceable ciphertext pronounceable would be just filling in "filler-vowels" according to a pre-defined pattern.

The problem here is that the VMS vowels aren't thrown in at random.  They have certain places within words where they can go, so you would need to explain the existence of those "vowel slots", which is going to be basically the same as accounting for the vowels themselves.

Yet another technique would be separately enciphering consonants and vowels of the plain text. I think that's the trick that would make any ciphertext pronounceable. Consider an example which is neither substitution, nor verbose. Don't know if it is a known cipher, but I just borrow it from the differential encoding used in the telecom world. Here, each subsequent character is not enciphered per se, but what is enciphered is rather the difference between it and the preceding character. E.g., the English alphabet runs: A, B, C, D etc. Let's assume we encode each line from scratch and the first word of the line is the word "odd". "O" is the first character of the line. It is enciphered just by its number in the alphabet - i.e., "15". The second letter is "D", which is 16 letters forward from "O" ("O" being counted itself). Thus "D" is enciphered with number "16". The third letter "D" is one letter distant from the second letter "D" (because we count the letter itself), hence it is coded with "1". We could well measure the distance beginning with 0, not with 1, but beginning with 1 is convenient for subsequent mapping of numbers to cipherext letters. In our example, 15 is "O" in the English alphabet, 16 is P, and 1 is A. So we get "OPA" as the ciphertext for "ODD".

Applied to the whole alphabet, such ciphertext would be unpronounceable for long phrases. What would make it pronounceable is considering two numbered rows - one for consonants (B, C, D, F...), another for vowels (A, E, I, O...) and enciphering consonants of the plain text with the first row and vowels of the plain text with the second row, using the same differential encoding as explained above. This way the phrase "this is a cipher", if I made no mistake, will be represented as "TNIM AB U KINSYL", which is quite pronounceable.

This is clever, although without thinking about it too much it's not clear that it would be fully invertible once you split the consonants and vowels into separate rows, since you would then have more numbers than letters within each row.  In any case though, you're basically preserving the consonant/vowel distinction found in the plaintext here, so this example actually sort of reinforces my larger point that it's hard to see the apparent consonant/vowel distinction made in the VMS text as meaningless.  (And further considerations of word structure and things like that would of course show that the VMS text was not actually produced by this method.)