In trade finance, the scariest state isn't failed — it's pending with no updates. A SWIFT message goes out, the counterparty goes quiet, and nobody owns the silence. Applicant assumes beneficiary is slow; beneficiary never got pinged. Days evaporate.

The fix I built at Newgen was a Tracer Service: a watcher over the pending pile that notifies applicant and beneficiary when a transaction sits too long without movement. Conceptually a dead-man's switch for correspondence — the absence of an event becomes the event.

It sat alongside the Correspondence Event superstructure mapping SWIFT N-series messages to workflow events, so tracer alerts carried real context (which message, which leg, how long stuck) instead of generic "please check" noise. Alerts without context get filtered; alerts with a message reference and age get acted on.

Generalizing: every async pipeline needs an owner for silence. Retries handle failure; only a tracer handles quiet.