Because it’s superior; if we’re doing a refcount+drop style, we’ll probably end up re-reading files relatively often, which seems wasteful