AWS S3 us-east-1 — typo took out half the internet

Casualty stat: ~4 hours down · S3, EC2, console, half the web

An engineer running a debugging playbook fat-fingered a parameter. A command intended to take a few servers offline took down a much larger set, including subsystems that powered S3's indexing in us-east-1. The cascade was visible across half the consumer web for hours.