Explore the nuanced approach to handling Kubernetes node failures in this conference talk. Learn why immediate deletion of failed nodes may hinder root cause analysis and prevention of future issues. Discover an alternative strategy that balances proper failover with the need for thorough investigation. Examine how existing projects handle node failures and gain insights into a proposed implementation leveraging External Remediation in MachineHealthCheck and fencing technologies like fence_kdump. Understand the importance of preserving failed nodes as valuable sources of information for engineers investigating system issues in cloud native environments.