Serverless jobs make it easy to run production notebook workloads at scale, but until now, the state of a run was gone once it ended. If a pipeline ran for hours and failed on the last step, the variables, DataFrames, and other intermediate state you...
