A meta-analysis wouldn't solve the problem unless it somehow included previously unpublished data. Publication bias has left whole fields of science untrustable.

If an individual scientist cherry-picked interesting data the way publications cherry-pick results to publish, it would be called fraud.