Rely on cache when deciding whether to backup a repo?
Our backup code uses project.empty_repo? to determine whether it should tar up the bare repo. I'm not suggesting this has been a problem in the past but the recent issue with caching in the RC make me wonder - are we willing to stake a backup on a cache?
Is there a reason we shouldn't just go ahead and tar up the bare repo even if it's empty? There's obviously some overhead related to time and probably a very small amount in disk space, but is it worth it?