Microsoft has explained what caused last week’s global Microsoft 365 outage, which left millions of people, including councils, government agencies, courts and schools in Australia, without Outlook email for most of Friday 13 November. The short version: a software bug that had been sitting quietly in Microsoft’s network servers since July was triggered by a busy day, a routine traffic switch and some servers already being offline, and it then spread across the world in under two hours.
The explanation comes from a preliminary post-incident report Microsoft released to customers on Monday 16 November (Australian time), and from an abridged version posted on X by Charles Lamanna, Microsoft’s executive vice-president of Copilot, Agents and Platform. Here is what Microsoft says went wrong, what it still hasn’t explained, and what Australian organisations can take from it. We covered the outage itself as it happened in our report on 13 November.
What Microsoft says caused the outage
Microsoft blames three things that, it says, were harmless on their own but together took down the “networking gateway”, the front door that sits in front of many Microsoft 365 services, including Exchange Online (the service behind Outlook email).
- A latent bug from July. According to Lamanna, a fix prepared specifically for Microsoft 365’s networking gateway servers introduced “lock contention inside an internal process which caused the machine to become unavailable”. In plain terms, too many tasks were queuing for the same shared resource until a server stopped responding. The change started rolling out in July and, Microsoft says, ran worldwide for more than three months without problems. Lamanna says “neither the underlying defect nor the fix applies to broader Windows deployments”, so this was not a bug in the Windows on your PC.
- A surge in load. Microsoft says it runs two separate gateway systems. About 15 hours before the outage, Teams failed over from the alternative gateway onto the main one that also carries Exchange Online, which increased the load and changed its traffic patterns. Microsoft describes this kind of failover as part of normal operations.
- Reduced capacity. At the same time, Microsoft says, “nearly 14% of server capacity for the networking gateway was out of service due to an unrelated issue from the previous day.” It says the service still had enough headroom to cope with that alone.
Together, Microsoft says, the extra load and reduced capacity tripped the July bug during the peak of US traffic. As servers stopped responding, Microsoft’s own traffic management moved users onto other servers, which then hit the same problem. “In less than two hours, the previously unknown defect was encountered across our fleet and rendered them unresponsive. At its lowest, more than 80% of servers were unresponsive,” Lamanna wrote.
Why it spread worldwide, and why Outlook took so long to return
Microsoft’s network gateway is split into thousands of “capacity units” grouped into regions, which are meant to act as bulkheads that stop a failure in one place spreading everywhere. Lamanna says that, “due to a separate issue”, those bulkheads did not stop the cascade. Microsoft hasn’t yet said what that separate issue was.
It also explains why some services came back long before email. Teams, OneDrive, SharePoint and Office apps were able to move off the faulty gateway, Lamanna says, “But not all services, like Exchange Online, can fail out of this network gateway.” He says Teams chats and meetings kept working for most of the outage, while Outlook email did not.
Recovery was slow for two reasons, according to Microsoft. Servers brought back online were “immediately overwhelmed by waiting traffic and became unresponsive again”, and because the defect was months old, “our normal rollback, flighting, and feature flags were unable to quickly revert the change.” Engineers instead built an update that put the code back to its earlier state and applied it to large blocks of servers at a time, in some cases more than half a region.
Why didn’t testing catch it? Microsoft says the problem is worse on newer machines with more processor cores, while “our load testing tended to focus on our weaker hardware SKUs with fewer cores.”
The timeline in Australian time
Microsoft describes the incident as starting on Thursday 12 November, which is US time. BleepingComputer reports it began at 11.20am US Eastern time that Thursday, which was about 3.20am AEDT on Friday 13 November, and that Microsoft confirmed services were fully restored at 11.10am Eastern on Friday, or about 3.10am AEDT on Saturday 14 November. That is why Australians woke on Friday to a full working day without email.
The Register, which has seen the preliminary report, says the Teams failover happened at 01:07 UTC on 12 November (12.07pm AEDT that Thursday) and that Microsoft identified and applied the fix that reverted the defective change at 07:20 UTC on 13 November, which was 6.20pm AEDT on the Friday. iTnews reported at the time that recovery for parts of Microsoft’s Asia Pacific infrastructure began at about 10pm AEDT on Friday night.
iTnews also reported that Home Affairs Minister Tony Burke said the government’s digital resilience agreement with Microsoft “has, in fact, been activated today to make sure that we get appropriate prioritisation of Australian interests in this.” iTnews noted that this agreement may have influenced Microsoft’s decision to prioritise Asia Pacific recovery, but Microsoft hasn’t said so.
What is confirmed, and what isn’t yet
The cause is Microsoft’s own account. It has been reported consistently by iTnews, The Guardian, ZDNET, BleepingComputer and The Register, but the full preliminary report sits in the Microsoft 365 admin centre, and no independent investigation has been published. Microsoft has said there was no cyberattack: according to iTnews, it said on Saturday that “this was a technical issue not related to cyber or AI [or] agents.”
A few points are still open. Microsoft hasn’t described the “separate issue” that let the failure cross regions, or the “unrelated issue” that took nearly 14% of gateway capacity offline. Its explanation has also become more specific over time: on Saturday it told customers the incident “was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change”, before Monday’s report pointed to the July fix. Lamanna says a final post-incident review “will provide more details later this week”, and The Register reports it is due on 20 November.
What Australian organisations should do now
- Read the report yourself. Microsoft’s Microsoft 365 Status account says the preliminary report is available in the admin centre under incident IDs MO1550821 and EX1550543. If you manage a Microsoft 365 tenant, or pay an IT provider who does, it is worth reading and filing with your incident records.
- Watch for fake outage emails. iTnews reported that Microsoft Threat Intelligence saw phishing emails impersonating Microsoft 365 service health notices and referencing the real Exchange Online incident EX1550543, falsely claiming held emails would be deleted within 24 hours unless recipients clicked a link. Treat any message like that with suspicion and check service health in the admin centre directly rather than clicking links.
- Check for missing mail. Microsoft extended its email timeout window to 48 hours during the outage to protect delayed messages, iTnews reported, and iTnews found some of its own scheduled emails arrived hours out of order. If something important was sent to you on 13 November, confirm it arrived.
- Plan for the next one. Make sure staff know how to reach each other and customers if email is down (phone lists, a non-Microsoft messaging channel, and a status page or social account). This outage, like the September Optus outage we covered, is a reminder to test those fallbacks before you need them.
Our read
Credit where it’s due: Microsoft has published a detailed and fairly candid explanation within days, including an admission that its load testing didn’t reflect the hardware it actually runs. But the uncomfortable takeaway is that this wasn’t an exotic failure. A three-month-old change, an ordinary failover and some servers already offline were enough to take email down worldwide for most of a day, and the safeguards meant to keep a regional problem regional didn’t work. Until the final report explains that “separate issue”, organisations should assume it could happen again. For Australian businesses and agencies that run almost everything through one provider, the lesson isn’t to panic and switch; it’s to make sure you can still communicate, and still find out what’s happening, when that provider has a very bad day.
Featured image: Servers inside a Microsoft datacentre in Boulder, Colorado, used by the Bing Maps team, 2010. Photo: Robert Scoble, via Wikimedia Commons, licensed CC BY 2.0; resized and converted to PNG. Illustrative only; this is not the networking gateway infrastructure involved in the outage.




