That’s very interesting. Typically a Rabin fingerprint is used to identify identical chunks of data.

Identifying similar blocks and, maybe sub-rechunking isn’t something I’ve ever considered.