Kubernetes Pod is restarting but logs are empty; how should the on-call engineer troubleshoot this?

I ran into a frustrating situation while on call last night: a service was repeatedly restarting in Kubernetes, but kubectl logs only showed empty output. My colleagues kept asking if the application wasn't logging anything, but in reality, the pod couldn't even get its main process running stably. I first checked the kubectl describe pod Events and found it wasn't an application error, but…