7.55Interview question: "How do you handle Salesforce outages?"

Strong answer:

“I'd classify the failure first. For transient connectivity/timeouts I'd use bounded retries with appropriate delay/backoff, ideally around a narrow idempotent operation. If retries are exhausted, I'd propagate or queue the work depending on whether the API is synchronous. I wouldn't retry validation, permissions, or schema errors. I'd also preserve correlation IDs and monitor retry exhaustion.”

Excellent.