How do you set up observability, tracing, and logging for a complex multi-agent system at Rvo Health to debug a failed run?
