* add additional info during exception of eks runs. * ensure we refresh k8s config after long calls. Kubernetes client on EKS has a 10 minute token time to live, so will result in unauthorized errors if tokens are not refreshed on long jobs.