Kubernetes Liveness Probes and CPU Limit Risks Self-Reinforcing CrashLoopBackOff

(Author here.) I thought this was an interesting problem we ran into, especially in light of yesterday's "Kubernetes is hard" discussion https://news.ycombinator.com/item?id=35331887 .

It's a failure mode that's not exclusive to Kubernetes, but can happen in production anytime there's a self-monitoring/self-restarting feedback loop with potential for a backlog to build.