Curious how one glitch took out so many States
Yes it is, isn't it?
I am a computer developer by trade. I am here to assure you that no upgrades, no backups, no 'drills', no 'testing' like this happens on active-users-on-system time. These sort of things occur only on off-hours. All the explanations offered so far DO NOT MAKE ANY SENSE to this computer developer with over 20 years in the field.
Furthermore, such a large system is NOT centralized. A good data architecture would have many redundant copies in distributed servers, synched using normal synching tools. I've personally created much smaller systems with this redundancy and synching. Have you ever heard of an entire large portion of a banking system going down? An entire branch of a phone system? No, you have not. That is because any large and complex system has many different branches or nodes of operability. The concept that one mistake brings it all down -- or even a lot of it -- is impossible.