Consider using Casa (or even ZFS) for Haddock storage

There are a *lot* of duplicate Haddock files generated, since each snapshot gets its own copy of every file.

If the files were stored in content-addressed storage, it would reduce the number of writes per snapshot, the amount of storage used, and the related expenses and disaster recovery considerations.

I mention ZFS because it supports transparent deduplication and compression. The server already uses ZFS; simply writing the files to disk in a correctly-configured dataset might be easy to implement.