What happens if a third-party integration is unavailable?

Short explanation:

An unavailable integration should not cause unlimited waiting or silently lose accepted work. The design needs deadlines and an explicit user-visible outcome, such as failure, pending work or a stale-but-labeled result. Retry and reconciliation policies depend on whether the external operation is safe to repeat.

Key points:

• Use a timeout and record the failure or pending state.

• Retry safe work later; show a clear status instead of waiting forever.

• Store enough context to continue later, and alert when queued work exceeds the acceptable delay.

Interview point:

• Explain what happens to the user's request while the provider is down, not only when the provider returns.