Adding a Workflow State Requires Auditing Every Downstream Projection

A recent engineering change introduced a soft cancellation state to an operational workflow, requiring updates across eleven files and sixteen integration tests. While storing the new state was simple, the real challenge was determining how it should be interpreted by every queue, counter, aggregate, snapshot, and export that reads the same record. Engineers used a state-impact matrix to explicitly decide whether each consumer should include, exclude, relocate, or annotate cancelled records. The state machine also needed clear rules on entry, exit, idempotency, restore paths, and how late-arriving competing events should be handled. The key lesson is that a workflow state is a system-wide contract, not just a data field, and leaving any projection's behaviour implicit risks silent contradictions across the system.