As last week fire demonstrated, it may happen that both a host and its backup can be unavailable at the same time. Even if that is only for a few days it is inconvenient and should not happen.
In the past few days backups for libvirt were implemented to send qcow2 images to OpenStack images so they can be retrieved if the libvirt hypervisor is destroyed. The next step will be to implement the opposite: downloading OpenStack backups (volume snaphosts and instance backup images) to a local host. Since those backups are a last resort, only the latest will be downloaded. If older data needs to be restored, in case someone accidentally destroyed data, the daily OpenStack backups can be used.