I think that if the offset was available publicly I could parse messages that have become concatenated on a stream like:

[1, 2, 3][4, 5, 6]

By using the offset to split the original byte array. Otherwise I can't think of a sensible way of parsing that. If you are ok with that, and the concept is correct, I am happy to make a PR. For now I will just copy-paste and modify the code.