Only rerun failed containers, not whole job


#1

Sometimes tests flake. Imagine we run a job with 10 containers, then why do I need to rerun all 10 if only one of them failed?


#2