The advent of web services brings new promises for organizations that collaborate amongst themselves through interorganizational workflow (IOW) management systems by providing a standard platform for interorganizational information exchange. However, a major impediment to more wide-scale use of such systems is the lack of provably correct methods for managing fault tolerance. The objective of our research is to provide a checkpointing scheme as the foundation for a sound recovery strategy for IOWs: a strategy that can be included in higher-level web services standards. We adopt concepts from the mobile computing literature to decompose workflows into mobile agent-driven processes that will prospectively attach to web services-based organizational 'docking stations.' This decomposition is extended in order to define logical points within the dynamics of the entire workflow execution that provide for locating accurate and consistent states of the system for recovery in case of a failure. Finally, we establish that the proposed scheme ensures consistency.