Microsoft has closed the incident that blocked mail to Outlook.com, Hotmail and Live for three days in November 2026 — an update to its spam enforcement rules, now fully rolled back. The rejected messages were permanent failures, so what is left is a recovery task, not a DNS task.
Microsoft closed the incident on 10 November 2026, and its own record of what happened is not ambiguous. Incident EX1545193 — "Some users may be intermittently unable to send email messages to Outlook.com, Hotmail.com, or Live.com email addresses" — ran from Friday 6 November 2026 at 18:00 UTC to Monday 9 November 2026 at 11:02 UTC, is classified as an incident against Exchange Online under "E-Mail timely delivery", and states its root cause in a single sentence: "A recent update to the spam enforcement rules caused email messages to be incorrectly marked and blocked, which led to impact."
That is a receiver-side bug, not a rule. It was not a tightening of the authentication requirements, not an extension of the high-volume requirement to ordinary senders, and not anything the rejected senders had done. The same bulletin says the update was "fully rolled back" and mail flow restored, then adds the line that matters for what happens next: Microsoft is "reviewing our update procedures and improving our mail detection logic to lower the false positive rate and ensure that legitimate email messages aren't inaccurately blocked."
Two days after closing the incident, Microsoft put a statement on the record with Windows Latest — a report headlined for the part senders will care about, that Microsoft confirmed its spam rule wrongly blocked them, and that senders were blamed for it.
"We've resolved an issue that may have prevented some email messages from reaching Outlook.com, Hotmail and Live addresses, and mail flow has been restored. We're continuing to monitor to help ensure email is delivered as expected."
Microsoft statement to Windows Latest, 12 November 2026The same report carries the two operational facts Microsoft gave alongside that statement. Customers "do not have to change your settings", which is to say the DNS work rushed through over the weekend was not the fix. And the messages already rejected will not be delivered late: the rollback restored the ability to receive new mail, not the mail previously refused. Our piece of 9 November, written while the block was live, left one question open — whether the stricter behaviour was intended and merely under-described, or a regression about to be reverted. It has been answered.
The window is worth keeping exactly, because it is the range you will need to search. Microsoft records the start as Friday 6 November at 18:00 UTC and the end as Monday 9 November at 11:02 UTC, with a final entry on 10 November closing the event. Windows Latest reports bounces in its own testing from at least 5 November, slightly ahead of that recorded start, and the reports collected on the mailop mailing list by Spam Resource's Al Iverson begin on Friday night US time.
Iverson's summary adds the two scope details most often got wrong while this was live. The block applied to Microsoft's consumer mailboxes — outlook.com, hotmail.com, live.com and msn.com — and not to Microsoft 365-hosted business domains, and it caught ordinary senders rather than bulk mailers. The rejections were intermittent rather than total, which is precisely what makes a receiver regression so easy to misread as your own problem: a resend may succeed, so the evidence looks like noise instead of a fault.
The bounce text made the misreading worse. Microsoft's incident bulletins describe the non-delivery reports as reporting "550 5.7.1 or 550 5.7.23 due to Service Provider Foundation (SPF) violation" — SPF there meaning Sender Policy Framework, though the bulletin expands the acronym incorrectly. The fuller text senders saw named their own records as the fault: 550 5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records. Windows Latest reports that support replies pointed senders at their own SPF configuration, and that senders whose authentication records were valid were caught anyway. The NDR was not a diagnosis; it was the receiver's fault describing itself as the sender's.
A 550 is a permanent failure. The receiving server has refused the message and told the sender not to try again, so nothing was queued, nothing retried after Monday, and nothing will arrive late. That is the operational consequence of this incident, and it splits the work into two halves with different owners. The technical half — the enforcement update — belonged to Microsoft, and it is closed. The half still open belongs to whoever sent mail into that window: a list of messages that were supposed to reach a customer between Friday and Monday and never did.
What typically disappears in a three-day refusal window is the mail nobody complains about immediately. Verification links and password resets, where the customer sees a form that does nothing. Order and payment confirmations, where the customer assumes the transaction did not complete. Support replies, where a case looks ignored. And any lifecycle message doing real work in the interval — a cart reminder, a renewal notice, a delivery update.
The search is mechanical. Filter the bounce log for 550 5.7.1 and 550 5.7.23, destinations on outlook.com, hotmail.com, live.com and msn.com, sent between 6 November 18:00 UTC and 9 November 11:02 UTC. Then ask a question most teams forget, because it concerns their own tooling rather than Microsoft's: what did the sending platform do with those addresses on the strength of a permanent failure? A hard bounce is what suppression lists are built to act on. If these 550s were recorded as hard failures, those addresses are now blocked for future mail, and the next campaign will skip exactly the customers this incident was supposed to recover — silently, which is the part that hurts.
The durable lesson of 6 to 9 November is a diagnostic one, and it is worth writing into the runbook while the example is fresh. Four checks separate a fault in your records from a fault in theirs.
_dmarc for the sending domain — one DNS lookup. Presence or absence is the fact to establish before touching anything else. If the record is there and mail to every other provider is arriving, the receiver is the suspect.None of this retires the real requirement, and the difference between the two is what belongs in your documentation. The requirement Microsoft publishes applies to domains sending more than 5,000 messages a day to its consumer services: SPF must pass for the sending domain, DKIM must pass, and a DMARC record must be present and align with SPF or DKIM — at minimum p=none, preferably both. Non-compliance is rejected with 550 5.7.515 Access denied, sending domain <domain> does not meet the required authentication level, and the guidance points at the domain in the 5322.From address as the one that has to authenticate.
Iverson's read of the episode is that senders who had already ensured compliance would have avoided much of the pain here. That is an argument for publishing the record. It is not an argument that the record was the problem. If your post-mortem says "Microsoft tightened DMARC enforcement in November 2026", the post-mortem is wrong, and it will send the next person hunting for a policy change the company has explicitly declined to make. The accurate sentence is narrower: the documented high-volume requirement asks for a DMARC record; an enforcement bug between 6 and 9 November punished its absence for three days; Microsoft rolled the bug back and says no settings need changing.
Read Microsoft's stated next step literally and the direction of travel is visible. The update was rolled back, not abandoned — the company says it is improving detection logic "to lower the false positive rate". The high-volume requirement predates the incident, so the rollback is a correction to enforcement rather than a withdrawal of the rules. Sending without a DMARC record is still a bet against a requirement Microsoft has published and has not retracted; the only thing this week proved is that the enforcement missed its target for three days.
_dmarc exists for every domain you actually send from — including subdomains, and the legacy domain still sitting in one template's From line.Microsoft's rollback restored the future, not the past. Every message refused between Friday 6 November and Monday 9 November was rejected permanently, so the part of this incident still waiting on you is a resend list — and a note that the DNS you changed over the weekend was never the problem.
Written by the team at LMR — a senior lifecycle marketing studio working in Iterable, Braze, Klaviyo and the tools you already run.
Start a project ↗