Retry Promise with Exponential Backoff

Retry with exponential backoff is how flaky networks stop becoming flaky UX — wait longer between attempts instead of hammering. Naive move Immediate tight loop retries, or recurse without a budget. Where it breaks 4xx errors should often not retry. Sleep must happen between failures, multiplying delay by factor each time. Exhausted retries must throw the last error. The fix Loop with attempt counter. try/await fn; on success return. On failure, if attempts remain and shouldRetry(error), await sleep(delay), then delay *= factor. Otherwise throw. Remember: Budget + predicate + backoff. Not every error deserves another try. Say this in the interview: I'd retry in a loop with a remaining budget, sleep with exponential delay only when shouldRetry says so, and throw the last error when attempts are exhausted.