For the love of god stop using CPU limits in Kubernetes

A Kubernetes analysis demonstrates that CPU limits significantly degrade application performance and increase costs by causing frequent throttling even when nodes have available CPU capacity. The study recommends removing CPU limits entirely while retaining CPU requests (which guarantee resource allocation) and memory limits, finding that this change can improve tail latency under traffic spikes, reduce startup times by 2x, and save tens of thousands of dollars per cluster annually through better resource utilization.