We are assuming that the all meta_sha256s are the same track. If these have different data_sha256s then we should cluster them on insert. If we have an mbid in one submission we should match the cluster to that mbid.
If 2 items in the same cluster have different mbids then for now we shouldn't put them in the same cluster.

If we have the same mbid but different meta_sha256 then we need to work out why. Possibly a spelling change in musicbrainz or a manually tagged file. It's a way of grouping different meta_sha256s. I'm not sure if this is something that should be done automatically or manually.