A checkout service began timing out for 2% of requests, with no clear pattern. Users saw 'payment failed' messages, yet logs across the API Gateway, Auth, Cart, Payment, and Notification Services all reported success. This classic distributed systems nightmare required a systematic approach to isolate the fault.

The developer used a trace ID assigned at the gateway to follow a single request's path. By grepping for a specific ID across all service logs, they discovered the Cart Service was returning a 200 status with an empty body. The Payment Service interpreted this as 'no items' and silently skipped the transaction.

This incident underscores the critical role of distributed tracing in complex microservice architectures. Tools like Jaeger or OpenTelemetry visualize these trace paths, making it easier to pinpoint failures. The key is propagating trace IDs consistently and logging them with every entry, turning a chaotic search into a targeted investigation.