Skip to the stories
AI & ML

Microsoft 365 Outage: How a Hidden Bug, Teams Traffic and Reduced Capacity Combined

  • John Martinez
  • 432 reads

A July gateway bug, a Teams traffic shift and 14% less server capacity converged, turning a recoverable fault into a prolonged Microsoft 365 outage.

Microsoft 365 Outage: How a Hidden Bug, Teams Traffic and Reduced Capacity Combined
Microsoft Outlook logo with DownDetector website opened in the background

Microsoft 365’s prolonged outage came down to three conditions lining up: a hidden networking bug, a shift in Microsoft Teams traffic and reduced gateway capacity. Each factor could be managed on its own, according to Microsoft. Together, they pushed servers into a failure state that was difficult to recover from.

The disruption affected services including Outlook and lasted roughly 24 hours across two business days. Microsoft’s explanation, shared with customers and on X, shows how a defect that had sat unnoticed for months became a system wide problem only when traffic patterns and available capacity changed at the same time.

A bug that surfaced only under load

The initial defect entered Microsoft’s networking gateway systems in a July 2026 update. The update was intended to fix an issue with Microsoft 365 gateway servers, but it also introduced a latent bug, Microsoft executive Charles Lamanna explained in a post on X.

Under certain load conditions, the defect caused servers to become overwhelmed and stop responding. Microsoft’s testing had not exposed it: during normal traffic levels, the affected systems behaved as expected. That left the bug dormant until conditions shifted enough to trigger it.

This is a difficult class of failure to catch with routine testing. A service can appear healthy under ordinary demand while still containing a fault that becomes visible only when traffic, routing and spare capacity combine in an unusual way. The account from Microsoft points to that interaction as the key to understanding why the defect remained hidden for months.

Teams traffic moved onto the affected gateway

Microsoft maintains two independent network gateway systems for Microsoft 365, a design intended to provide an alternative when one system has trouble. But about 15 hours before the wider disruption, Teams failed over from the alternative gateway to the primary gateway. The primary system also supports Exchange Online, the cloud service behind Outlook.

The traffic shift increased demand on the primary gateway. Under ordinary conditions, a failover of this kind would be part of the system’s resilience plan. Here, it sent more load toward servers that contained the dormant defect. The failover did not create the bug; it helped create the conditions in which the bug could cause trouble.

Less capacity left little room for error

At the same time, nearly 14 percent of the networking gateway’s server capacity was out of service because of an unrelated issue from the previous day, Lamanna said. Microsoft described the reduced capacity as routine maintenance, with enough servers normally available to keep service running.

That spare capacity mattered once the Teams traffic moved and the defect was activated. Increased demand met a smaller pool of available servers. Requests began to accumulate, and servers became unresponsive during a peak traffic window in the United States. With fewer working servers to handle incoming requests, the queue could grow faster than the system could clear it.

The combination is the practical lesson for engineers operating large services: redundancy and capacity reserves help, but their value depends on how they behave together. A second gateway can provide a route around trouble, yet a traffic shift can put pressure on another system. Planned maintenance can be safe at normal load, but it reduces the margin available when an unexpected fault appears.

Why restoring service took time

Recovery proved harder because bringing servers back online did not immediately clear the backlog. Waiting traffic overwhelmed them again, causing another round of unresponsive behavior. Microsoft also could not simply reverse the most recent software change: the defect had been present since July, so rolling back the latest update would not remove it.

Instead, the company had to develop and apply a patch to return the affected code to a previous working state. That made recovery a live repair under pressure, while incoming requests continued to arrive. Microsoft’s full explanation describes the interaction and repair process in more detail. Enterprise customers can also consult the Microsoft 365 Admin Center.

For customers, the account clarifies why an outage can persist after engineers identify the broad cause: restoring capacity and correcting code are only part of the job when waiting requests can overwhelm recovering systems. Microsoft has apologized, and affected businesses can submit a request for service credits.

The next useful signal will be how Microsoft addresses the conditions that let the defect escape testing and how it manages recovery when queues have already built up. The incident shows that service resilience depends not only on backup systems, but also on the headroom and repair paths available when several safeguards are under pressure at once.

Source: John Martinez · mashable.com

From the same set