How Microsoft Is Bringing Microsoft 365 Back: Rollbacks, Extra Capacity and Careful Traffic Control After 17+ Hours of Exchange Online Failures

Microsoft is reverting a network change, rebuilding server capacity and throttling traffic to end the EX1550543 outage, and reported its first signs of lasting recovery at 11:44 Mauritius time (07:44 UTC) on 13 November, but still has no fix time.
Where Things Stand
As of about 13:30 Mauritius time (09:30 UTC) on Friday, 13 November 2026, the Microsoft 365 outage that began at 20:17 Mauritius time (16:17 UTC) on Thursday is still not officially over. Microsoft has not marked either incident, EX1550543 (Exchange Online) or MO1550821 (the wider Microsoft 365 advisory), as "Service restored", and it has not given an estimated time for full recovery.
There is, however, real progress. At 11:44 Mauritius time (07:44 UTC) Microsoft reported "positive results for service stability and increased capacity" from its latest traffic optimisation work, and said it was rolling those actions out more broadly across affected regions to reach a persistent recovery. That is the first time since the outage began that Microsoft has talked about recovery that lasts, rather than recovery that collapses again under load. Some users are reporting that Outlook loads again, although slowly.
Microsoft has also said repeatedly that there is no indication of malicious activity. This is an internal capacity failure, not a cyberattack.
This article explains what Microsoft is actually doing to bring the service back, step by step, and why it has taken so long. All times are Mauritius time (UTC+4), with UTC in brackets.
What Broke, in Short
Microsoft has not published a final root cause, but its own updates describe a clear chain of events:
- A recent network configuration change lines up with the start of the outage. Microsoft first said it had found no matching change, then at 05:06 Mauritius time (01:06 UTC) on Friday said it had identified one and began reverting it.
- On the affected Exchange Online servers, the process that sets up secure connections became overwhelmed, so the servers stopped answering requests.
- Automated health checks saw those servers failing and removed them from service. The servers that were left took on their traffic, became overloaded too, and the failure spread.
- Later, the work to rebalance traffic caused a surge of calls from Microsoft's routing layer to backend servers, which knocked recently recovered servers back over.
Everything Microsoft is doing now is aimed at breaking that chain: stop more servers falling out, remove the trigger, rebuild capacity, and let traffic back in slowly enough that it does not knock the service down again.
Step 1: Stopping the Damage From Spreading
The first job was to stop the outage getting worse.
- Deployment freeze. At about 23:06 Mauritius time (19:06 UTC) on Thursday, Microsoft paused all deployments and configuration changes across the affected service, so nothing new could add to the problem while engineers investigated.
- Stopping servers being pulled. Once it saw that failing health checks were removing servers, Microsoft took steps to stop additional servers being taken out of service and tested a way to bypass the faulty health check.
- A new health check. At about 00:56 Mauritius time (20:56 UTC), it validated an alternative health-check method on a subset of servers, which let some of them return to service, and prepared an automated process to detect and restart unresponsive machines.
These steps limited the spread, but they did not fix the underlying fault. Servers that came back kept failing again.
Step 2: Undoing the Trigger
With simple restarts not holding, Microsoft turned to the changes that might have caused the failure.
- At 01:40 Mauritius time on Friday (21:40 UTC Thursday) it said it was assessing whether reverting a small number of recent configuration updates could speed up recovery.
- Around 04:00 Mauritius time (midnight UTC), it began rebuilding a subset of servers on a previous configuration.
- At 05:06 Mauritius time (01:06 UTC) on Friday, it said a network configuration change correlates with the start of impact and began reverting it.
- By about 08:56 Mauritius time (04:56 UTC), it was forcing machines to sync the recent network change, so that every server picks up the corrected settings and none is left running the bad configuration.
Rolling back a change across a global fleet is slow. Every server has to receive the fix, restart cleanly and prove it is healthy before it can take real traffic again.
Step 3: Rebuilding Capacity
Because so many servers dropped out at once, Microsoft also needs more working capacity than usual to absorb the backlog of users trying to reconnect.
- Restarting in isolation. At about 06:00 Mauritius time (02:00 UTC), Microsoft reported positive results from restarting affected infrastructure and recovering servers in an isolated manner, bringing small groups back and watching them before adding more.
- Adding capacity. At about 06:30 Mauritius time (02:30 UTC), it began applying capacity improvements to support regions as they came back online.
- Scaling up virtual machines. Around 08:56 Mauritius time (04:56 UTC), it said it was scaling up VMs and repairing machines stuck in a bad state.
- Recovering the frontend. At 10:12 Mauritius time (06:12 UTC), it said it was continuing to recover its frontend servers, the layer that first receives connections from Outlook and other apps.
Step 4: Controlling the Traffic
The hardest part of this recovery has been traffic. Every Outlook client, phone and Teams app in the affected regions keeps retrying, so as soon as servers come back they are hit with a wave of reconnections.
To manage that, Microsoft has:
- Redirected traffic away from the worst-affected infrastructure, starting with US traffic on Thursday afternoon, then spreading Outlook traffic across its worldwide infrastructure.
- Applied protective traffic limits in some locations and limited some traffic types by location.
- Adjusted traffic protocols, redistributed traffic and applied "strategic traffic blocking" (about 08:36 Mauritius time (04:36 UTC)), which produced partial recovery in the regions where it was used.
- Reviewed traffic flows and routing to find where traffic management had become unbalanced.
At 11:44 Mauritius time (07:44 UTC), Microsoft said these latest traffic optimisation actions were producing better stability and more capacity, and it began applying them across all affected regions.
Why Recovery Keeps Stalling
The outage has now passed 17 hours because each fix has run into the same wall: servers come back, get flooded, and fall over again.
At 06:54 Mauritius time (02:54 UTC), Microsoft admitted its changes were "not taking as expected" and said it faced "sustained traffic management imbalances". Its later explanation was that the rebalancing work itself caused a surge in outbound calls from routing services to backend servers.
This is a familiar pattern in large outages, sometimes called a thundering herd: millions of clients retry at the same moment, and the rush of reconnections becomes a second outage. The only reliable fix is to bring capacity back gradually, region by region, while holding traffic back just enough that recovered servers can stay up. That is what Microsoft's latest updates describe.
What "Restored" Will Look Like
When the service does come back, it is unlikely to be a single moment:
- Region by region. Expect some regions and some users to recover before others, as Microsoft widens the traffic changes it started at 11:44 Mauritius time (07:44 UTC).
- Slow at first. Outlook may be sluggish, and some apps may need restarting before they reconnect.
- Delayed email. Mail that queued during the outage will keep arriving for a while after the incident is closed, and some messages may arrive out of order.
- Official end point. The outage is only over when Microsoft marks EX1550543 and MO1550821 as "Service restored" in the Microsoft 365 admin center. That time is what counts for SLA claims.
- The report. Microsoft normally publishes a preliminary post-incident review in the admin center within a few days, with a final report later. It should confirm which change triggered the failure and why the health checks and traffic rebalancing made it worse.
What You Should Do While It Recovers
- Track the incidents EX1550543 and MO1550821 under Service health in the Microsoft 365 admin center.
- Warn users about phishing. Expect fake "Microsoft 365 restored, sign in again" emails and fake IT support calls. Microsoft will not ask users to sign in through an email link to restore a mailbox.
- Do not bulk-restart everything at once. If you manage many devices, let Outlook and Teams reconnect naturally rather than forcing every client to reconnect at the same moment.
- Check what arrived late. After recovery, review message trace, quarantine and delivery logs for mail delayed or unchecked during the outage, and confirm Intune check-ins have resumed.
- Keep records of the timeline and business impact for SLA credit claims and your own incident report.
Reef Sentinel will keep following the recovery and update its coverage when Microsoft confirms the service is fully restored.
