The Day the Internet Coughed (Again)
Today’s outage is a reminder that platforms without redundancy, single-provider dependencies and untested recovery plans aren’t just commercial risks, they’re moral ones when people’s work relies on what we build. The web was born decentralized, but our stacks drifted the other way. It’s time to design for failure on purpose.
Days like today should reset our appetite for risk. When a single hyperscaler stumbles, the ripple turns into a tide. In the early hours of October 20th, a fault centered in AWS’s US-EAST-1 region cascaded across the web. Reports pointed to DynamoDB API problems and, for a period, even DNS resolution issues for that endpoint. Error rates and latency spread to other services in the region. The blast radius was significantly bigger than expected. Consumer apps, banks, and enterprise tooling blinked out, with major outlets tracking the roll-up as it unfolded. It’s a familiar pattern by now. When US-EAST-1 coughs, the internet catches a cold, but the scale this time made the dependence hard to ignore.
If you build on SaaS, this is your wake-up call. A platform without redundancy, a stack tied to one region, a “DR plan” no one has tested, those are bets you’re placing with other people’s livelihoods. Most of us aren’t writing code that decides life or death, but payroll, point-of-sale, logistics, banking access, healthcare portals and the everyday scaffolding of work do run on these rails. When they stall, trust erodes, revenue slips, and suddenly the abstract “platform risk” is a very concrete conversation with your team about missed targets and missed rent. You don’t just have a commercial obligation to harden what you build, you have a moral one, too. The internet’s original spirit was decentralisation and end-to-end resilience, but our industry’s consolidation has concentrated failure. We can’t rewind history, yet we can design like the network will break, because it will.
From a European vantage point, the dependency is starker. Our public sector, our mid-market, our startups, so much of it rides on US-owned infrastructure governed by US law. There’s movement to change that. The Commission is funding sovereign cloud procurement and clarifying what “sovereign” should mean in practice. Providers like AWS are preparing a European Sovereign Cloud operated only by EU personnel and competitive pressure is shifting incentives around portability and fees, with Google dropping certain EU data-transfer charges to grease multicloud paths ahead of the Data Act’s provisions. None of this flips the table overnight, but it signals a direction, more European control, clearer guardrails, more room to choose without lock-in.
What to do with that reality is the part that matters. Design for graceful degradation before you design for heroics. Treat regions and providers as commodities where you can, and as specific risks where you can’t. Common patterns still work. Keep customer-visible paths stateless and cache-friendly so you can swing traffic between regions or CDNs, keep data models ready for change with append-only logs and replication strategies that tolerate lag, keep contracts explicit so failover targets can be validated by test. Make health signals first-class, automate the switchover, and rehearse it. Regularly. If complete provider independence is out of reach, reduce the single points of failure you control. Run active-active across at least two regions, use provider-neutral DNS and observability, isolate critical queues and identity paths from the rest of the blast radius and document the manual play you’ll run when automation refuses to cooperate.
Multicloud deserves a grown-up conversation. It isn’t free and it isn’t a religion. It’s insurance you buy where the exposure warrants it. Revenue-critical edges, regulatory choke points, places where a regional crack can become an existential break. The ability to re-route traffic, drain state, and bring up a parallel path,whether across regions of the same provider or between providers, turns an industry-wide incident into a degraded day, not a ruined quarter. Europe’s push for sovereignty may add tailwinds here. More interoperable offerings, clearer portability rights and market pressure that rewards designs which can move. That was always the internet’s promise: resilience through distribution, autonomy through choice. Today was a reminder to build like we mean it.
I’ve been the architect, sat beside the architect and cleaned up after the architect. Everything here was learned the hard way and in production.
