If the source document doesn't exist, the MultiWorkflow usecase thinks it is invoked on an area node (e.g., /authoring/) and creates a pre-order enumeration of all nodes in the area starting with the top-level nodes. This is not clean and causes a performance penalty.