[oooyay]

I suspect on GitHubs front this has to do with how they populate their status site. They may update it manually once they identify customer impact. If they're using internal metrics to qualify the status site then they're likely not using all of the needed metrics to reflect customer impact. There's also a third possibility which is that between you and GitHub there's something that causes a partition or failure that is outside of GitHub and your domain of control.

I agree with you that the ultimate value is in customer impact. I was saying "that's hard" but synthetic monitoring is not the solution because it doesn't achieve what it sounds like it achieves.