It sounds like that's a possibility, but why on earth would you take the time to setup a 3 node cluster of object storage for reliability and ignore one of the key tenants of what makes it reliable?