Compare repeated trials with a decision trail.
Recent revisions are kept. Restoring one creates a new revision.
