I download still-running fiction from the internet and read them locally, and periodically replace the files with versions with more chapters. It does not sound like the workflow would work for me.
(This also means hashing the contents wouldn't work.)
But if the annotations get stored directly in the ebook file (as opposed to separately in a JSON file), I don't see how this would work for you, either? You would still have to transfer them to the new file somehow.
(This also means hashing the contents wouldn't work.)