I wonder if any lessons were learned here? No reference in article to other potential fixes involving waiting until async code finishes. Seems like an obvious one (in hindsight)