@Andriy Knysh (Cloud Posse) github.com/cloudposse/atmos/pull/3040 since we run on spot instances, and run quite a lot of jobs, there's some % that get interrupted simply by infrastructure issues. This is introduces a job that runs when workflows complete and inspects why they failed; if it is related to losing connections to the github control plane, instances getting shutdown, etc, it will retrigger them automatically.