On btrfs, ansible playbooks can be run on a chroot snapshot under eatmydata, and the snapshot can replace the main chroot in the end.
This would allow a much faster update time, and if maintenance is transactional (see #2), eatmydata's lack of consistency becomes irrelevant
