That's an awful analogy because "realistically you will be able to do all the things you want to do". If a random GitHub service goes down there's a significant chance it breaks your workflow. It's not always but it's far from zero.

One bulb in the cluster going out is like a single server at GitHub going down, not a whole service.