One reliability report counted 2,730 outages across 34 AI and LLM providers in the first half of 2026, about fifteen a day, and a model suspension that ran 18 days 19 hours, taking the platforms built on top of it down with it. A second vendor is not a second failure domain. These three checks take an afternoon; the fourth is a recurring five minutes. Nothing here needs a migration.
The two things that make a fallback fake
# 1. shared dependency
primary -> model family M, region R, gateway G
backup -> model family M, region R, gateway G # same hour, same outage
# 2. shared budget
client: { timeout: 8s, retries: 3 } # one pool for both providers
^ the slow primary spends it before the backup is ever calledThe checks
Everything starts ticked. Untick what you will not do; the choice is saved on this device only, and Print keeps only what you kept.
Why. Resellers, gateways and clouds collapse into the same upstream more often than the contract suggests: the same model family, served from the same region, behind the same gateway. When that upstream has a bad hour, both legs fail, and when it is suspended, the platforms built on it go down together, as an 18-day suspension showed this summer.
The check.
- For each provider write down three things: whose model weights, which cloud and region, which gateway or proxy.
- If any two match the primary, that leg is not a fallback. It is a second route to the same failure.
Source: H1 2026 Cloud and SaaS Reliability Report, read 1 Oct 2026
Why. A single client-side budget is spent by whoever is slow. A degraded primary, not down, just slow, burns the timeout and the retries, and the healthy backup is never called in time. Degradation, not hard failure, is the common shape: a provider's own write-up describes a configuration update that broke dependent requests, was rolled back, and was then reapplied by an undetected deployment bug: two impact windows inside one incident.
The check.
- Per provider: its own timeout, its own retry count, its own circuit breaker.
- Cap total attempts across the ladder.
- Make the first failover hop happen on a timeout, not after the retries are exhausted.
Source: A major provider's own incident write-up, 25 Jul 2026
Why. Staging proves the code path, not the dependency graph: staging has different keys, different quotas, different regions and no real traffic. Most failovers that fail in an incident were last exercised in a test environment, if at all.
The check.
- Put a five-minute window in the calendar.
- Flip the primary off in production configuration, watch error rate, p95 and cost, then flip it back.
- Write down what the backup's latency and spend were. That number is your failover budget.
Why. When both providers are impaired, the only options left are the ones you built earlier: a cached answer, a smaller local model, or a deterministic response that is honest about being one.
The check.
- Name the degraded mode per feature, and the signal that turns it on.
- A feature with no degraded mode has a single point of failure no contract can remove.
The outage count and the suspension length are the report's figures; the incident shape in check 2 is from the provider's own write-up. Nothing here is a vendor claim.
Take this list with youThe PDF is the handout: this list on one or two pages. The Markdown is for your own docs, your wiki, or an assistant you want to hand the list to. Print gives you this list only, without the navigation.
Something in your production AI broke in a way that is not on this list? Tell me what broke.