Agreed, having separated databases just for that purpose is a bad idea as people could be doing work in the wrong database without knowing it and the practice necessitates duplicating all the prior year entries with no discernible benefit.

The backup is really only there if you somehow manage to completely lose everything in the live dataset and need a reference, or need to start the year end over again (something you wouldn't consider after a day or two has passed). It wouldn't ever get unzipped under normal circumstances.