Do you see same patterns?

[Anton]

OK, now I understand. You mean that the author used a substitution cipher, and then introduced a set of transformations such as duplication to obfuscate the text, introducing more characters, making it sort of verbose cipher. By following these transformations in the reverse direction we may disclose the original substitution.

Conceptually that makes sense, although it's not clear what to do with labels (many of them may disappear from the text with this approach), also this will not explain some line- and paragraph-level phenomena, such as multipass or prevalence of gallows in paragraph-starting positions.