Sulfur dioxide has a pretty short lifespan in the stratosphere. If an experiment does go "horribly," and honestly I can't imagine how an experiment, as opposed to a full scale geoengineering project, could really go that badly, it should be easy enough to just stop.