Reliability

A case study on a self-chaining Map/Reduce that reconciled CSV imports by searching for the imported journals, and used up its retry budget in about a minute. The fix: retry for a fixed amount of wall-clock time, not a fixed number of attempts.

A case study on NO_DEPLOYMENTS_AVAILABLE and RCRD_HAS_BEEN_CHANGED: how a workflow action that queued a Map/Reduce could undo an approval during bulk runs, and how an eventually consistent sweep fixed it.

Why NetSuite integrations hit 429 Too Many Requests, how to size a client-side pool, retry with backoff and jitter, and use external ID upserts so a retry never creates a second sales order. With a live concurrency simulator.