> Most of these sites are not even that big. I expect maybe a few thousand visitors per month.

> This demonstrates again a simple fact: if you put your site behind a centralized service, then this service is a single point of failure. Even large established companies make mistakes and can go down.

I'm guessing sites with a few thousand visitors a month don't much care about single points of failure. Seems like kind of a circular argument - if they're too small to care about needing a proxy in front of their service, then they are also probably too small to care about the handful of events that cause it to go down every so often.

People talk about "single points of failure" like invoking that phrase in and of itself means something is bad. There are many areas where avoiding single points of failure is essentially impossible. It's about how much risk and impact you are willing to tolerate with those points of failure.