Recently, with the AWS outage, our stack of loads of different cloud providers ended up working pretty well! It might be a bit complex running distributed nodes and updating state via API, but its cheap and clearly resilient.