Plan study sample sizes with uncertainty in view.
Recent revisions are kept. Restoring one creates a new revision.