Let us count the days

[Jorge_Stolfi]

Looks promising; there is a 2020 paper by Matlach, Dostal and Janeckova  (VM: Symbol Roles Revised) which basically argues that the entire VM is a cipher

That was also my You are not allowed to view links. Register or Login to view. for why the distribution of the lengths of words in the lexicon matches very closely the binomial function choose(n,k).

But currently I think that this theory is unlikely.  For one thing, before computers, a codebook cipher would be a viable encryption method only for short and highly sensitive messages.  I think it would be very hard to write a whole book that way.  And even harder to read it. (Although the Rohonc example may kill this argument...)  

But mainly because I found a better explanation for that graphical coincidence, that also explains may other puzzles of the VMS.

All the best, --stolfi