If the program in the worker pods completes or fails very fast, there will be pending and then succeeded/failed k8s events. And the running event will be missing. For this scenario, we add the pod state transfer logic from pending to succeeded/failed.
If the program in the worker pods completes or fails very fast, there will be pending and then succeeded/failed k8s events. And the running event will be missing. For this scenario, we add the pod state transfer logic from pending to succeeded/failed.