I looked at a very similar thing (briefly) a long time ago. The reordering I tried involved sorting the characters alphabetically in each word, but also alternating the consonants and vowels.
Example: "r e o r d e r i n g" --> "e e i o" and "d g n r r r" --> "d e g e n i r o r r"
The main purpose was to reduce the character bigram entropy. This worked to some extent, but not nearly enough to achieve the low levels of Voynichese. Note that I did this long before Stolfi's word paradigms came out, and these strongly suggest that such a type of word game is not likely to be sufficient to 'explain' Voynichese.
Of course, I was also worried whether this operation was reversible, because it creates collisions, i.e. several plain text words could be mapped onto the same encoded word. I tried that out using an English text, and it turned out that such collisions are quite rare. Much more rare than I had thought.
Example: "r e o r d e r i n g" --> "e e i o" and "d g n r r r" --> "d e g e n i r o r r"
The main purpose was to reduce the character bigram entropy. This worked to some extent, but not nearly enough to achieve the low levels of Voynichese. Note that I did this long before Stolfi's word paradigms came out, and these strongly suggest that such a type of word game is not likely to be sufficient to 'explain' Voynichese.
Of course, I was also worried whether this operation was reversible, because it creates collisions, i.e. several plain text words could be mapped onto the same encoded word. I tried that out using an English text, and it turned out that such collisions are quite rare. Much more rare than I had thought.