Node.js Error Tracking: Failing Closed During Pricing Rule Rollouts

The article argues that an uncaught exception in a Node.js service should be treated as a process-fatal event, not something to log and keep serving through. The recommended response is to record the error synchronously, stop accepting new work, and let a supervisor replace the process. For a logistics pricing rule gated by a feature flag, this gives rollback a clear boundary, since the flag can revert to the previous rule while the unhealthy worker drains or exits. The author contrasts this with routine request-level errors, which can be handled with ordinary try/catch, and notes that Node.js documentation warns against resuming normal operation after an uncaught exception. The piece stresses that error records should carry rule version, flag revision, and shutdown reason so operators can identify the last known-good rule from a single event.