API Retry Budget

A practical example

Example: during a two-hour CRM outage, your connector would normally retry each of 9,000 failed pushes indefinitely. The retry budget allows three attempts per event and a global ceiling of 20,000 retry calls per incident.

Once hit, remaining events move to the replay queue, preserving quota for the first live syncs after recovery.

What to evaluate before investing

• Ask whether the cap is configurable per event, per job, and globally, and where the threshold is set.
• Verify what happens when the budget is exhausted: deferral to a queue, alerting, or silent abandonment.
• Check whether retry consumption is reported separately from primary sync usage.

Limitations and tradeoffs

A tight budget may abandon events that would have succeeded on a later attempt, while a loose one still risks quota exhaustion; the right ceiling depends on your outage tolerance and replay capability.