How can Facebook have an entire system outage when they have datacenters all over the world?

I do devops and programming, and I'm not too versed with networking outside the realm of devops. I setup load balancers often. I attended a Facebook conference where I learned they do adopt the devops philosophy at Facebook. And this is why I can't understand how a huge system like Facebook could have a major outage when they have servers all over the world.

In a well designed network architecture how is something like this possible? Asking because I'd like to learn how these things happen.

Thanks.