Session traces and cost controls are emerging as key observability techniques for diagnosing AI agent failures, helping teams spot tool-call loops and runaway spend while preserving enough execution context for post-incident debugging.

Pinterest Engineering cut Apache Spark out-of-memory failures by 96% using improved observability, configuration tuning, and automatic memory retries. Staged rollout, dashboards, and proactive memory adjustments stabilized data pipelines, reduced manual intervention, and lowered operational overhead across tens of thousands of daily jobs.