Hi. We have a workflow that failed over the weekend, due to the return type in a task not matching the task annotation. Instead of just failing, it said it failed in the console, but also said that it was running, and the pod stayed live until we saw it today. Has anyone seen anything like that, where a pod gets stuck running? It can be tough when GPUs are involved, due to the expenses, so we're wanting to figure out how to avoid that.

Hey Andrew Was it a mapTask?

It was not. A regular task, but under a conditional. Oh and under a sub workflow

WHAT~!!!!!!!

@freezing-airport-6809 oops I was thinking of the wrong workflow, its actually just a direct task call, no sub workflow and no conditional. I'll DM you some snippets if that's ok