“Twitter has two data centers running that can handle the entire site being failed into it.”
Each data center runs at no more than half capacity, so either one can carry the whole site. The cache team got about one page a week. Twitter didn’t fall over when half the staff left because the people who left built it not to. Automation buys time. It doesn’t fix the things that break slowly while nobody’s watching.