Handling failures in software workflows, particularly at scale, is a complex challenge for developers, as disruptions like software bugs or service outages can affect thousands of customers. To address this, a new primitive called "fork" is proposed, allowing workflows to be restarted from a specific step and code version, effectively "rewinding time" to recover from failures without repeating previous successful steps. This method is illustrated through scenarios like payment service outages and localization bugs in invoice generation, where "fork" enables developers to resume processes from the point of failure, ensuring timely and context-aware recovery. The implementation of "fork" in a workflow engine, such as DBOS, involves checkpointing completed steps in a database, allowing workflows to be resumed efficiently and programmatically. This approach enhances the robustness of workflow management, making it easier to handle complex failures and ensuring that business processes remain resilient even when issues arise.