Catch bad data before it breaks your pipelines.
Recent revisions are kept. Restoring one creates a new revision.
