f17r multispectral images

[oshfdk]

I would suspect that in pen tests one would normally try to repeat some shape or word a few times? Voynichese marginalia looks mostly like a phrase or a charm with only a few repeating elements.

I would rather believe that at least part of the marginalia are somehow related to the key or mnemonics used to en/decipher the manuscript. For example, if the cipher involves a table of Latin characters, the marginalia could be used to reconstruct this table based on some principle known to the author(s). This would make sense in case author(s) believed the original key or table could be lost.

Recipe-like words in the marginalia would just be a method of obfuscation or a consequence of the active vocabulary of the author/scribe. For example, the author(s) got the encoding table, and wrote a short sequence where every third (or fifth or whichever) letter is part of the table and the rest are just random letters added to create some resemblance of meaning. If you are dealing with recipes, the random words you conjure up when trying to fill in the pattern created by the cipher table would be "so nim" and "gasmish", etc.