That was a neat blog post! Can you clarify this bit: "If a pipeline fails, the associated MR is kicked from the train and all preceding pipelines are restarted.

That was a neat blog post! Can you clarify this bit: "If a pipeline fails, the associated MR is kicked from the train and all preceding pipelines are restarted." - Do you mean all the pipelines _already started_ are restarted? Because I'm guessing that if a pipeline fails, the preceding pipelines shouldn't be affected, strictly speaking.