We\'re using Dask to distribute compute work across an EMR cluster. We\'re using Dask-Yarn. We\'ve noticed that when we experience node failures sometimes those failures will ta
