What Actually Breaks When a Site Goes Viral Overnight
It's rarely the server. It's the assumptions baked into the code that nobody thought to question.
Everyone assumes a viral spike takes down the server. It almost never does. Cloud infrastructure scales horizontally now, and most hosting providers can absorb 10x traffic without you lifting a finger. The server is the least of your problems.
What actually breaks is everything you built assuming traffic would stay small and predictable. The contact form that emails you directly instead of queuing. The image upload that resizes synchronously on the request thread. The analytics script that makes a blocking call to a third party with no timeout. None of these matter at normal traffic. All of them fall over in the first ten minutes of a spike.
The second thing that breaks is your database, but not in the way people picture. It's rarely raw load. It's a handful of unindexed queries that were fine at 50 rows and became catastrophic at 50,000, or a connection pool sized for a dev environment that never got revisited. A single N+1 query pattern, invisible for months, becomes the entire outage.
Then there's the human layer, which breaks just as hard as the technical one. Nobody knows who has access to the DNS provider. Nobody remembers the password to the payment gateway dashboard. The one engineer who understands the deploy pipeline is asleep in a different timezone. Traffic spikes are a technical event and an organizational fire drill at the same time, and most teams have only prepared for one of those.
The teams that survive virality well aren't the ones with the fanciest infrastructure. They're the ones who already asked "what happens at 100x" before it happened, even if the honest answer at the time was "we don't fully know." That question changes what you build, not because you over-engineer everything, but because you stop writing code that quietly assumes the world stays small.