Why one cloud outage takes down half the Internet

On 20 October 2025, one list of addresses in Northern Virginia went blank, and by breakfast in London, Ring doorbells, Snapchat, Fortnite and the apps of big British banks had stopped. Nobody had typed a thing. Here is how one fault at AWS travels that far: cloud regions and availability zones, a regional service and its DNS name, a race condition, a cascading failure and hidden dependencies. Then what keeps an app standing when its region falls, and what that costs.