The git data model isn't ideal though, it misses content-defined chunking of file content and directory entries, which leads to lots of duplicate data with large text files or directories containing large numbers of files. Newer backup tools like restic/borg support this though.
That seems like an implementation detail, not a fundamental design decision as it should be easy to change how packfiles are implemented. I'm not sure it would be an improvement though: it already only stores deltas for similar objects.
