It Wasn't Hackers. It Was a Router in Arkansas.

Everybody calm down
Over the last two weeks, Microsoft 365 went down, the FAA grounded every domestic flight, Chase got slow, United got stuck, iCloud and Apple Pay took a nap, and the NSA lost its own internal network. If you were on social media, you know exactly who did it. It was China. Or Russia. Or a teenager in a hoodie. Or all three, working together in a dark room with green text scrolling down the walls.
It was none of them. It was us. It's almost always us.
The official list of suspects
Here's what the companies themselves say broke:
- Microsoft 365 (Nov 12): A "latent defect" from a network capacity upgrade done months earlier. It sat there quietly until a later traffic shift woke it up. Microsoft says it was "not related to cyber or AI / agents," which is a sentence nobody would have needed to write five years ago.
- The FAA (Nov 13): A nationwide ground stop that Spectrum News reports was traced to a single router in Arkansas. One router. Thousands of flights. I've had home routers do this to my Netflix, but never to Delta.
- Workday (Nov 13–14): A disaster recovery environment caused a disaster. That's a little like your smoke detector starting the fire.
- Chase (Nov 16): A software bug in on-premises systems.
- Apple (Nov 17): An internal update.
- The NSA (Nov 17): Scheduled maintenance. The White House said, "We see no evidence that this was a cyberattack," which is reassuring, coming from the people who'd be the first to know if the NSA got hacked. Or the second, after the NSA.
To be fair, real attacks did happen this month. Hackers abused a signup flaw at Optimizely and planted malicious JavaScript that Report URI tied to 178 organizations' sites. Hospitals, libraries, schools, and at least one town in Maine got hit with the usual ransomware-flavored misery. But the big, scary, everything-is-down outages? Those were self-inflicted.
The call is coming from inside the data center
Anyone who has worked in software for more than a week knows the most dangerous thing in any system isn't a hacker. It's a change. A config push, a capacity upgrade, a "quick" maintenance window, an update that passed every test in staging because staging has the traffic of a rural Blockbuster.
Microsoft's outage is the textbook case. The bug shipped months earlier and did nothing. That's what latent means: it waits. Then some unrelated change shifted traffic, and the old bug finally had the load it needed to ruin everyone's Thursday. No single change was the problem. Two harmless changes, months apart, were the problem.
A few lessons, none of them new, all of them apparently still necessary:
- Blast radius matters more than bug count. You will ship bugs. The question is whether one of them takes out a region, a country, or the entire domestic air travel system.
- Your recovery system is part of your production system. If the disaster recovery environment can corrupt production data, it's not a safety net. It's another trapeze.
- Single points of failure are still out there. Somewhere in Arkansas, a router is getting a lot of attention it didn't ask for.
- Test under real load, not polite load. Staging is where bugs go to hide.
The hackers showed up anyway
Here's my favorite part. The day after Microsoft's outage, attackers sent fake Microsoft 365 "service health" emails that used the real incident ID. The pitch was that the emails held during the outage would be deleted within 24 hours unless you clicked a link.
So the hackers didn't cause the outage. They just waited for it, like a guy selling umbrellas outside the subway when it starts raining. That's the real security lesson of the month. Attackers don't need to break your systems when your systems break on their own and everyone's already confused and clicking things.
So, should you panic?
No. But if you get an email this week saying your outage-delayed messages are about to be deleted, maybe don't click it. And if you work in IT, go hug your change management process. It's been through a lot.
As for the cause of the next big outage, I'll make a prediction right now. It won't be a foreign government. It'll be a Tuesday afternoon config change that somebody described in the ticket as "low risk."
chris@nearauth.ai