You're arguing that on shoring QA would reduce the probability of something going wrong. I'm neither going to agree nor disagree.
However, I think the failure here is to mitigate the impact of something going wrong. Their rollout plan was fundamentally flawed - it shouldn't have taken out so many machines at the same time. It should have been rolled out in stages, with only 1 machine at most at any given customer receiving early versions.
It's best to assume a bug will get through 1 day or another, and spend some time mitigating the other axes too.