When a customer workflow breaks, the cause is usually buried somewhere in a long chain of calls to outside services we don't control. We can't tell which hop failed or how it affects the customer. A good solution must follow a single request through external integrations and tie reliability targets to customer workflows, not just servers.