# Fallbacks and graceful degradation interview questions > 6 fallbacks and graceful degradation interview questions for AI and ML engineer roles, each with a model answer. Canonical: https://llmdojo.dev/interview/topic/fallbacks Last updated: 2026-10-23 6 questions: 1 junior, 3 mid-level and 2 senior. Taught in [Fallbacks and Graceful Degradation](https://llmdojo.dev/learn/production-ai-systems/fallbacks-and-degradation). ## [What is graceful degradation?](https://llmdojo.dev/interview/q/what-is-graceful-degradation) Level: Junior When parts fail or load is too high, return **lower-quality, cheaper results** instead of nothing: a smaller model, no reranker, a cached answer, the raw passages, or an honest message. Never a hang. **Diagram: A ladder of fallbacks** 1. Normally, the primary model and the full pipeline answer. 2. If it's down or rate-limited, a gateway can route to another model. AI Engineering notes API failures are unfortunately common. 3. Under load, do less: the SRE book's graceful degradation returns lower-quality, cheaper-to-compute results rather than nothing. 4. With no model at all, a cached answer or the raw retrieved passages may still help. 5. The bottom rung is an honest message or a person. What's unacceptable is a hang, a blank screen, or an error the user can't act on. ## [Your primary is 99% available and a fallback 99.5%. What's combined availability?](https://llmdojo.dev/interview/q/your-primary-is-99-available-and-a-fallback-99-5-what-s) Level: Mid-level If outages are **independent**, both are down 1% × 0.5% = 0.005% of the time: about **2 minutes** a month instead of about 432. If they share half their outages (same provider or region), it's about 217 minutes. Pick fallbacks by **how** they fail. ## [What does a circuit breaker do?](https://llmdojo.dev/interview/q/what-does-a-circuit-breaker-do) Level: Mid-level Counts failures on a call; past a threshold it **opens**, failing fast or going straight to the fallback with no timeout wait. After a cool-down it lets **one trial** through; success closes it. **Diagram: A circuit breaker** 1. Martin Fowler's circuit breaker wraps a remote call and counts its failures. 2. When failures reach a threshold, it trips: further calls return an error immediately, without reaching the failing service, so a fallback can answer at once. 3. After a timeout, it lets one trial call through to see whether the problem is fixed. 4. If the trial succeeds, the breaker resets; if not, it opens again. 5. Fowler's reasons: without it, callers tie up resources waiting on timeouts, failures cascade across systems, and the struggling service keeps getting hit. In the course's example, users wait 80 s in total instead of 1,000 s during an outage. ## [Demand is twice capacity. Why shed load instead of queueing everything?](https://llmdojo.dev/interview/q/demand-is-twice-capacity-why-shed-load-instead-of-queueing-everything) Level: Mid-level Queueing everything makes every request slow until most time out: nobody is served. **Load shedding** rejects some requests fast with a clear “busy”, so the rest are served well. The SRE book's example: return 503 to new requests once too many are in flight. ## [You fall back to a model you've never evaluated with your prompt. What can go wrong?](https://llmdojo.dev/interview/q/you-fall-back-to-a-model-you-ve-never-evaluated-with-your) Level: Senior It's a different model: format reliability, refusals and quality may all change, turning an outage into a quality incident. Give the fallback path its own prompt version if needed, its own evaluation run, and metrics split by which model answered. ## [Retry after a bad output, or send two requests in parallel?](https://llmdojo.dev/interview/q/retry-after-a-bad-output-or-send-two-requests-in-parallel) Level: Senior Depends which budget is tighter. Retrying after failure is cheap but can **double latency**. Parallel requests keep latency near one call but pay for redundant calls every time. Either way, write the failure behaviour down and **exercise it**: unused fallback paths rot. ## Related - [All interview questions](https://llmdojo.dev/interview)