Hi Brian, thanks for your response. I was referring specifically to parallelizing your Python code with Dask. But it's interesting that you achieved parallelism using SubDags - afaik SubDags are run by SequentialExecutor to prevent from deadlocks.