If it's one service though there is still always a single point of failure outside of your control. If you want true redundancy you need to use multiple versions with no interconnections and develop your own solution to mitigate partial failure, of course your own solution is also a single point of failure, but at least it's under your control.

That said, Github has never been down long enough for me to justify a fallback procedure and policy for my team. The complexity of the solution outweighs the benefit. We'll cross that bridge when we come to it.