From Thursday afternoon (UTC) until Friday morning, Microsoft’s Exchange Online service was largely unable to accept and deliver mail. As an email delivery provider, we send millions of messages a day to Microsoft-hosted mailboxes, so our queues work like a live sensor on Microsoft’s mail infrastructure.
Our queues to Microsoft-hosted domains grew very long as a result. They are now draining, and delivery is returning to normal. Below, we also share what our own delivery data showed, hour by hour.
What happened
The outage began around 16:17–17:45 UTC on Thursday, November 12, and service started to recover at about 11:20 UTC on Friday, November 13. Microsoft tracked it as incident EX1550543 (Exchange Online) and MO1550821 (other Microsoft 365 services that depend on it).
- Thursday, ~17:30 UTC: independent monitors (ThousandEyes, Kentik, Cloudflare) see Microsoft 365 connections failing worldwide.
- Thursday, 17:54 UTC: Microsoft’s first public update reports mailbox access problems in the Americas; it soon confirms any region may be affected.
- Thursday evening: Microsoft pauses deployments, restarts servers and reroutes traffic, but early fixes do not hold.
- Friday, 11:20 UTC: Exchange Online traffic starts to recover.
- Friday, ~14:00 UTC: Microsoft says email queues are close to normal levels.
What our queues saw
That daily volume turns our delivery data into a detailed record of how the incident unfolded. When Exchange Online slows down or stops accepting mail, we see it within minutes in our own delivery logs, often before an official status update.
On a normal day, almost every message we hand to Microsoft is accepted on the first try, in about 4 seconds. In the 17:00 UTC hour on Thursday, our retries to Microsoft started climbing, and by 18:00 UTC one message in seven needed retrying. That matches the onset reported by independent network monitors, and Microsoft’s own first public update at 17:54 UTC.
The outage then came in two waves. Retries peaked around 52% late on Thursday, eased overnight as Microsoft brought some servers back, then rose again to about 80% on Friday morning. For mail submitted at 09:00 UTC on Friday, the median message waited about 2.5 hours to be accepted; for the slowest 5% of Thursday evening’s mail, the wait was 10 to 13 hours.
Recovery shows just as clearly. From 12:00 UTC on Friday, the share of retried mail fell quickly, and by 16:00 UTC the median delivery time was back to about 6 seconds.
What it meant for your emails
When a receiving mail server can’t accept a message, the sending server doesn’t give up. It keeps the message in a queue and retries at intervals. That is exactly what happened with mail addressed to Microsoft-hosted domains: messages were held and retried, not rejected.
Because the outage lasted almost 18 hours, those queues grew far longer than usual. Many messages sent on Thursday afternoon and evening reached inboxes only on Friday morning. Microsoft also extended its own email timeout to 48 hours, so mail delayed inside its systems is kept rather than dropped.
The outage also hit the services around the inbox: Teams, SharePoint, Copilot and sign-in all had problems where they rely on Exchange Online. Recipients may simply not have been able to open their mail, even when it had arrived.
Where things stand now
Microsoft is accepting mail again and our queues are draining steadily. Of the roughly 3.6 million messages we have sent to Microsoft since Thursday, about 16,000 (under 0.5%) are still waiting for a retry. Once Exchange Online came back on Friday morning, its traffic briefly ran about 19% above normal as the backlog from across the internet arrived at once, according to network data from Kentik.
As of Friday afternoon (UTC), Microsoft reports sustained improvements but has not yet declared the incident fully resolved. You can follow the official status on Microsoft’s cloud service status page. Some recipients may still see short delays or intermittent access problems while the last of the backlog clears. We are monitoring delivery to Microsoft domains closely and will update this post if anything changes.
What you should do
In most cases, nothing: queued messages are being delivered automatically.
- Don’t resend. Resending campaigns or transactional emails will add to the backlog and may cause recipients to get duplicates.
- Expect odd numbers for Thursday’s sends. Delivery, open and click times for Microsoft recipients will be shifted to Friday, and opens may look low until the backlog clears.
- Check time-sensitive messages. One-time codes, password resets and expiring offers sent during the outage may have arrived too late to be useful. Consider a follow-up where that matters.
