top of page

Three Months Dormant, One Day Dark: What Microsoft's Outage Report Teaches Leaders

1 hour ago
5 min read

Editor's note: This post follows our November 12 analysis of the Microsoft 365 outage. It draws on Microsoft's preliminary post-incident report as reproduced and reported by technology news outlets. The report is preliminary. Microsoft has said a final version will follow by November 20, 2026.


Bottom line up front: Microsoft says a software defect that sat unnoticed since July caused last week's Microsoft 365 outage, which disrupted service for nearly 24 hours. The dormant bug makes the headline, but it is the least useful part of the story. The real lessons are that automated safety systems spread the failure, that warning signs built for about 15 hours, and that a long quiet period created confidence the facts did not support.


Every large computer system carries hidden flaws. Engineers know this. What separates a minor fault from a global outage is what happens around the flaw: how it is tested, how quickly it is noticed, and whether the system contains it. On those three measures, Microsoft's own account is more revealing than the headline.


What Microsoft Now Says Happened

Microsoft's report describes an internal software fault. It contains no reference to an intrusion. In plain terms, the company gives this account:

  • The flaw. In July, Microsoft began installing a software fix on its networking gateway servers. These are the "front door" machines that connect users to email and other services. The fix contained a defect. Under certain traffic conditions, the servers ran out of processing power and memory and stopped responding.

  • The quiet period. The fix ran worldwide for more than three months without visible trouble.

  • Three triggers. About 15 hours before the outage, an automated switch moved Microsoft Teams traffic onto the gateway that also serves email. Nearly 14 percent of gateway capacity was already offline because of an unrelated fault. Then the peak of the United States business day arrived.

  • The collapse. As servers failed, traffic moved to other servers, which failed in turn. Within two hours, more than 80 percent of the gateway servers were unresponsive.


Charles Lamanna, a Microsoft executive vice president, apologized publicly. "We are deeply sorry for the impact to all our customers," he wrote.


The Safety Net Spread the Failure

Large cloud services are built to route around trouble. When one server fails, traffic moves to a healthy one. That design works when failures are isolated. It works against the provider when every server carries the same flaw.


That is what happened here. Each redirected wave of traffic pushed another group of servers past the limit. Barriers meant to keep a failure inside one region also did not hold, so the disruption spread worldwide. Microsoft attributes that to a separate issue, which the preliminary report does not describe.

"The systems built to keep the service alive carried the failure from one server to the next."

Recovery was slow for a related reason. Restarted servers met a backlog of waiting traffic and went down again. Engineers eventually had to take servers out of service in large groups, repair them, and bring them back.


Fifteen Hours of Warning, and No Way Back

Two details deserve more attention than they have received.


First, the warning signs. The Teams traffic moved at about 8:07 p.m. Eastern time on Wednesday, November 11. Over the following hours, load rose and isolated servers failed in some regions. Automated systems handled those failures quietly. The monitor that tracks user connectivity did not raise its first flag until 11:03 a.m. Eastern on Thursday. Automation absorbed the early symptoms, and in doing so it hid the trend.


Second, the testing and the rollback. Microsoft says its laboratory tests focused on less powerful machines, while the defect is worse on newer hardware with more processing cores. The test did not match the real environment. Because the change was months old, the usual tools for reversing a bad update could not undo it quickly.

"Three quiet months did not prove the change was safe. They proved it had not yet met the wrong day."

What Our First Analysis Got Right, and Wrong

OSRS published its first assessment on November 12, while the incident was still developing. Readers deserve an honest accounting.

  • What held. We wrote that no evidence pointed to a cyberattack and that leaders should wait for the company's review. Microsoft's account supports that position.

  • What we corrected. We wrote that inboxes stopped loading shortly after noon Eastern. Microsoft's timeline places its first high-priority alert at 11:20 a.m. Eastern. We also relayed Microsoft's early statement that traffic had not increased. The report now names added Teams traffic as a trigger. The original post has been updated.

  • What needs a limit. We wrote that reliability is the provider's job and continuity is yours. That remains true. However, Teams, OneDrive, and SharePoint could move to alternative infrastructure, and Exchange Online could not. No customer setting would have restored email. Continuity therefore means a second channel that does not depend on the provider at all.


An Analyst's View

The strongest case for Microsoft should be stated plainly. The company published a detailed, self-critical timeline within days. Failures born of three unrelated conditions meeting at one moment are difficult to predict, and many organizations would have disclosed far less.


Nevertheless, three questions remain open. Why did the regional barriers fail? What took 14 percent of capacity offline the day before? Was an earlier Teams disruption that day connected? The final report should answer them.


The wider lesson reaches beyond Microsoft. A police dispatch center, a bank, a hospital, and a ministry all rely on systems with automatic failover and long-settled software. Each should ask the same questions of its own technology and of every provider:

  1. Do you test changes on the same class of hardware that runs in production?

  2. Can you reverse a change that is three months old, and how long does that take?

  3. Does your failover have limits that stop a failure from spreading?

  4. Do your monitors watch what users experience, or only what machines report?


How OSRS Can Help

OGUN Security Research and Strategic Consulting LLC serves an international audience that includes the intelligence community, law enforcement, academia, diplomats, and business leaders, and we present analysis to decision-makers in the United States and abroad. We can help your organization:

  • Review continuity plans and independent backup communications

  • Assess dependence on single technology providers

  • Prepare vendor questions on testing, failover, and recovery

  • Deliver intelligence briefings on technology risk for executives and boards


Share this article with a colleague who manages technology risk, and subscribe to our email list to receive our assessment of Microsoft's final report.


Enjoyed this article? Stay informed by following us on Google News, Twitter, and LinkedIn for more exclusive cybersecurity insights and expert analyses.


Sources: BleepingComputer; Office Watch (reproduces Microsoft's preliminary report); ABC7 News. Microsoft's report is published inside the Microsoft 365 admin center, and OSRS could not open it directly. Times were converted to Eastern time.


About the Author

Dr. Sunday Oludare Ogunlana, Founder and CEO of OGUN Security Research and Strategic Consulting LLC, Professor of Cybersecurity, national security scholar.


Intelligence. Protection. Strategy. www.ogunsecurity.com

Comments

Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page