The Belastingdienst had moved about 5,000 of 47,500 staff to Microsoft 365 before stopping; mail and calendars come in-house in 2027, with archive tooling still unresolved.
Microsoft blames a July gateway fix for the Microsoft 365 outage
EditorialBy TrustList Editorial
A change that ran without trouble for three months, a Teams failover and 14% of capacity already out combined on 12 November; the final review is promised this week.
- Cloud
- Incident Management
- Email Management
- Collaboration
About Microsoft blames a July gateway fix for the Microsoft 365 outage
Microsoft blames a July gateway fix for the Microsoft 365 outage
16 November 2026: Microsoft says the global Microsoft 365 outage on Thursday 12 November began with a defect in a fix it had been running on its networking gateway servers since July, which left more than 80% of those servers unresponsive at the worst point. Exchange Online took longest to return, and the company has promised its final post-incident review later this week.
Not yet independently verified. The account comes from a post on X by Microsoft executive Charles Lamanna, which carries no date on the page we read; iThome dates it to 16 November 2026. Microsoft's incident report sits in the Microsoft 365 admin centre, which only customers can open, and the final post-incident review has not yet been published. The post does not state how long the outage lasted. We will update this when it can be confirmed, and remove this note.
Three conditions that had to coincide
The explanation came from Charles Lamanna, a senior Microsoft executive, in a public post. He describes three factors.
The first was a latent defect: lock contention inside an internal process that made servers stop responding under load. It arrived with a fix prepared specifically for the gateway servers that sit in front of many Microsoft 365 services. That fix was first deployed in July 2026, reached the whole fleet over several weeks and ran for more than three months without incident. Lamanna writes that "neither the underlying defect nor the fix applies to broader Windows deployments".
The second was load. About fifteen hours before the outage, Teams had failed over from its alternative gateway to the primary gateway that also carries Exchange Online. The third was capacity: nearly 14% of gateway capacity was already out of service because of an unrelated problem the day before.
Why testing missed it and isolation failed
Load testing did not catch the defect because the test labs used a different mix of traffic and server types, on weaker hardware with fewer processor cores. The fault is worse on newer machines with more cores, which is what production runs.
During peak traffic in the United States, requests built up until servers stopped answering. Redirecting traffic away from failing servers spread the problem to healthy ones, and in under two hours the defect had reached the fleet. The partitions meant to keep a failure inside one unit of capacity did not hold, because of a separate issue, so the impact became global.
Exchange Online had nowhere to go
Teams, OneDrive, SharePoint and the Office apps can move away from a faulty gateway, and they recovered earlier; Lamanna says Teams chats and meetings worked for most of the outage. Exchange Online cannot fail out in the same way and came back more slowly.
Recovery was slow for a reason administrators will recognise. Servers that restarted were at once overwhelmed by the backlog. The usual tools, a rollback, staged flighting and feature flags, could not undo a change that was months old. Engineers instead built an update that returned the code path to a known good state, then took servers out in groups, half a region or more at a time, patched them and restarted them.
"We are deeply sorry for the impact to all our customers," Lamanna wrote.
Before the final review arrives
The incident record is in the Microsoft 365 admin centre under Service health, and tenant administrators should save a copy now, since service credit claims under Microsoft's online services agreement rest on the incident reference and the tenant's own record of downtime. The post says nothing about credits.
Two details bear on continuity planning. Mail was the last service back, so a plan that assumes email is the fallback channel when Teams fails had the order backwards on this occasion. And a single shared gateway tier carried several services at once, which is the design question the final review, due later this week, will need to answer.
Related on TrustList:
Sources
- Charles Lamanna (Microsoft) on X: account of the 12 November outage (read 17 November 2026), 17 November 2026
- iThome: report on the Microsoft 365 outage explanation, 17 November 2026
Categories & features
- Cloud
- Incident Management
- Email Management
- Collaboration
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
You may be interested in
More than 70,000 organisations turned the setting on in preview; it stays off by default for existing tenants, while new ones get it with no way to switch it off.
Existing Microsoft 365 tenants lose password-based SMTP sending by default at the end of December, though admins can switch it back on. Tenants created after December will not have it, and a final removal date comes in 2027.
Project Online retires on 30 September 2026 and its data becomes inaccessible. From 1 October, Microsoft starts deprecating EWS for cross-tenant free/busy, MailTips and calendar sharing, which move to Cross-Tenant Access Policy.
Kleiner Perkins, Redpoint and Khosla co-led the round, with Sequoia and Battery taking part, at a valuation the company puts at $1 billion. Traversal says monthly investigations grew from 513 to nearly 73,000 this year.
An attacker used OPSSHIELD's billing server to upload a fake cPGuard update that ran as root for about 70 minutes, collecting password hashes, SSH keys and MySQL root passwords. Update to 5.88.26 and rotate every secret.
Robusta, founded in 2021, rebuilt its Kubernetes troubleshooting tool around an AI agent; it is the second company Freshworks has agreed to buy in October, after Torii, and no price has been given.
Someone used credentials for Amadeus system accounts between August 2023 and September 2026. Asiana says booking reference numbers were taken and that Amadeus is still working out what else was.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More CloudThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.