Microsoft Exchange Online is Wrongly Locking Away Customer Mailboxes

Key points

• Microsoft has been fighting an Exchange Online fault, tracked as EX1436407, that has wrongly quarantined customer mailboxes since 19 July.
• Affected users cannot reliably send or receive email, and senders writing to them get bounce messages back.
• Microsoft blames a recent infrastructure change that chewed through memory and triggered the false quarantines.
• Cleanup of the excess indexing data reached 72% by Wednesday evening, up from 66% earlier that afternoon.
• The company has not said which regions or how many customers are hit, and has given no completion date.

Microsoft is still trying to fix an Exchange Online fault that has been locking customer mailboxes away by mistake since Sunday. Exchange Online is the cloud email service that businesses rent from Microsoft to run their staff inboxes and calendars.

The incident, which Microsoft tracks internally as EX1436407, started on 19 July. It was still not resolved as of the most recent update on Wednesday evening.

While the fault runs, some users cannot receive email at all. Others cannot send. Calendars have also been out of reach. People trying to email an affected mailbox get a Non-Delivery Report back, the automated bounce message that tells a sender their email never arrived.

What actually went wrong?

A recent change to Microsoft's own back-end infrastructure ate far more memory than expected, and the servers ran out. When that happened, the system wrongly flagged customer mailboxes as suspicious and quarantined them, meaning it walled them off from normal email flow.

"A recent infrastructure change caused excessive memory consumption from unexpected indexing data results in an out-of-memory condition, causing affected mailboxes to be incorrectly quarantined," Microsoft said in its advisory.

Indexing data is the behind-the-scenes catalogue Exchange builds so it can search your mail quickly. In this case, that catalogue ballooned and starved the servers of memory.

Microsoft says this is a repeat of an earlier problem it logged as EX1434354. The first fix, in other words, did not stick.

Who is affected?

Microsoft has not said. The company has not published a list of affected regions, and has not disclosed how many tenants or mailboxes are caught up in the fault. It has labelled the event an "incident," a classification it reserves for problems with noticeable user impact.

First reported by BleepingComputer, the outage is the latest in a string of Exchange Online mishaps where legitimate mail got wrongly held back.

How is Microsoft fixing it?

Engineers are working through the excess indexing data and clearing it out. Progress moved from 66% complete on Wednesday afternoon to 72% by Wednesday evening. As memory pressure eases in each region, Microsoft is pulling mailboxes back out of quarantine.

The company has not committed to a completion time. It said only that a timeline would come in a later update. The next scheduled update was set for 6 p.m. UTC.

What should ordinary users do?

If your work email has been patchy since Sunday, this is likely why, and there is nothing you personally need to fix. If you sent an important email and got a bounce message, resend it once services are stable. IT teams should watch Microsoft 365 admin centre notifications under incident ID EX1436407 for regional updates rather than opening new support tickets, which will only add noise.

A pattern of quarantine bugs

This is not the first time Exchange Online has grabbed the wrong mail. In March 2025, an anti-spam bug quarantined legitimate messages. In May 2025, a machine-learning filter (software that learns to spot junk mail from examples) started marking real Gmail messages as spam. A September fault blocked Exchange and Teams users from opening links. And in February, over-eager anti-phishing rules flagged thousands of clean URLs as credential theft.

Each was fixed. Each also cost customers hours of missing mail while Microsoft worked out what its own systems had done.