Insights
Universal Systems

By Chris Wall
5 min

What happened
I've been watching the outage news stack up over the last week and I think it's worth talking about. What happened to these organizations happens to mid-sized companies all the time. We just don't make the news.
Microsoft 365 started failing on the afternoon of Thursday the 12th. Outlook, Teams and SharePoint were down or unreliable for users in 74 countries, and service wasn't fully back until Saturday morning. Microsoft says the fault was introduced during an earlier network capacity upgrade and exposed by a later network change.
The next day the FAA grounded every departure in the country for a little over four hours. The cause they've given so far is routers on their telecom network overloading the air traffic control system. That network still runs on time-division multiplexing, and a review the FAA commissioned in 2023 called it obsolete. They've paused installing the routers until they and L3Harris work out the root cause.
Monday afternoon United's pilots couldn't get their takeoff numbers from dispatch. That's the weight and balance and performance data they need before they can leave the gate. Departures sat for about 90 minutes. United hasn't said what failed, but crews were told not to log out of their apps because they might not be able to get back in. That detail tells me more than the press statement did.
This morning it was the NSA. Their internal classified network went down, and people familiar with it are pointing to routine patching and software upgrades. They staggered arrival times at Fort Meade while they worked on it.
All four said early on there was no sign of a cyberattack. I get why that's the first thing everyone wants to hear. It doesn't change much for the people who couldn't do their jobs.
What actually went wrong
Most of these started with somebody doing their job. Microsoft was adding capacity. The NSA was patching. The FAA was in the middle of replacing equipment, and the old gear underneath couldn't keep up with what was being added on top of it. Secretary Duffy said it pretty directly: the legacy equipment went down and the new equipment didn't.
I see a smaller version of this all the time. There's a firewall or a core switch that's two years past end of support. It's still running, nobody can get a downtime window to replace it, and nobody wants to be the one who touches it. Then a firmware update or a new piece of hardware asks more of it than it can handle, and that's when you find out what was depending on it.
Modernizing is still the right call. The exposure is in the stretch between old and new, when the two are tied together and nobody has written down how.
What I'd be asking this week
Start with your end of life list. Not the one from the last audit. I mean every device and system that's out of support today, with a name next to each one for who owns replacing it. If that list doesn't exist, building it is the first project.
Then look at rollback. Almost every change ticket I review has the rollback section filled in, and very few of those rollbacks have ever been run. Pick one significant change this quarter and back it out on purpose so you know the process works.
Microsoft was down for close to two business days for some customers. Think through how your people would reach each other and your customers if Outlook and Teams were gone that long. A phone list saved in SharePoint doesn't help when SharePoint is the thing that's down.
You also need to know how quickly you can tell whether an outage is an attack. The White House pulled people into the Situation Room on Friday because nobody could rule it out right away. Your incident response plan should say who makes decisions in those first few hours before you have an answer, and what they're allowed to shut off.
Last, figure out what stops working when sign-in does. United telling pilots to stay logged in suggests authentication was somewhere in the failure path. Most organizations have put nearly everything behind Entra ID or another identity provider. That's the right move, but it means a bad day there takes down a lot more than it used to.
The next six weeks
Thanksgiving is next week and year end is right behind it. This is when teams try to clear the change backlog before people leave, and it's also when there are the fewest people around if something breaks.
I'm recommending a freeze on anything that isn't security related from Thanksgiving week through the first week of January. Patches for actively exploited vulnerabilities still go out, with a tested rollback and someone on call. Anything else that can't wait needs a named approver.
If you'd like us to go through your end of life inventory or your change process before the holidays, call us at 866-455-0493 or email tickets@usicomputer.com.
Sources
• WindowsForum: Microsoft 365 outage traced to network changes
• CryptoBriefing: Microsoft 365 outage hits 74 countries
• FAA: Nationwide Ground Stop Q&A
• ShiftDelete: FAA ground stop traced to aging routers
• Al Jazeera: A router switch forced the ground stop
• NBC Chicago: United outage blocked takeoff data
• CNN: United resolves system outage
← All insights
© 2026 Universal Systems, Inc. All rights reserved.