DMARC p=none in 10 minutes, and what the Outlook fault taught us
In early November 2026, a lot of perfectly ordinary email to @outlook.com, @hotmail.com and @live.com addresses bounced. Signup codes, password resets, invoices. Microsoft has since confirmed it was a fault on its side,
In early November 2026, a lot of perfectly ordinary email to @outlook.com, @hotmail.com and @live.com addresses bounced. Signup codes, password resets, invoices. Microsoft has since confirmed it was a fault on its side, rolled it back, and there is no new DMARC requirement.
So why write about it? Because the incident was a cheap lesson for anyone whose app sends email, and the defensive setup it pointed to takes about ten minutes. This post covers:
- What happened, according to Microsoft and the people who measured it
- What it taught us about how apps lose email
- The fix: check, publish
p=none, align - Where Spamhaus's new HELO/PTR rule fits
- Subdomains,
sp=, and alerting on the right bounce - Proving your mail actually arrives
What happened
Microsoft tracked it as incident EX1545193 in Microsoft 365 Service Health. The notice itself sits in the Microsoft 365 admin center, so you need an admin role to read it there. Public copies and press reports quote it. The final update reads:
"A recent update to the spam enforcement rules caused email messages to be incorrectly marked and blocked, which led to impact."
and:
"We've fully rolled back this update and confirmed that mail flow has been restored and impact is resolved."
Microsoft gives the window as "Friday, November 6, 2026, at 6:00 PM UTC" to "Monday, November 9, 2026, at 11:02 AM UTC" (mirror of the notice; also quoted by The Register). A Microsoft spokesperson told Expert Insights that mail flow had been restored and that senders don't need to change their settings.
A few details matter for developers:
- The bounce. The notice says users "may have received a Non-Delivery Report (NDR) reporting 550 5.7.1 or 550 5.7.23". Senders outside Microsoft 365 mostly saw this:
550 5.7.1 Message rejected: Message lacks SPF/DKIM/DMARC Auth Records (AS9031519)
Microsoft 365 senders got it wrapped in a 550 5.7.23 "SPF violation" NDR, which sent many admins off checking SPF records that were fine all along (Engagor).
- The timeline is a bit fuzzy at the start. Operators reported bounces before Microsoft's start time. MailRoute saw them in its logs on 5 November, and Maborak found a Microsoft Q&A report from 6 November at 09:05 UTC. Microsoft's notice says 6 November, 18:00 UTC.
- It was intermittent. One message bounced, the next got through.
- Consumer mailboxes only. In Spam Resource's test on 7 November, mail to outlook.com was refused while the same mail to a Microsoft 365-hosted business domain was accepted.
-
What it keyed on. Microsoft hasn't said. In tests that operators ran while the rule was live, domains with passing SPF and DKIM but no DMARC record were refused. In DMARKOFF's tests, as summarised by Maborak, adding a record, even
p=noneand even only on the parent domain, meant 0 of 112 messages were refused. That's other people's measurement, not Microsoft's explanation.
While it was happening, plenty of people read it as Microsoft quietly starting to require DMARC. The Hacker News thread was full of that theory. It turned out to be a bug. Microsoft's only published DMARC requirement is still the one for domains sending more than 5,000 messages a day.
What the fault taught us
1. Permanent bounces don't come back
A 550 is a permanent failure. Under SMTP, a 4xx gets queued and retried, but a 5xx means "don't try again". So when Microsoft rolled the rule back, the mail it had refused stayed refused. Nothing was redelivered.
For an app, that's the worst kind of failure.
- Nothing retries. The verification email is simply gone.
- Nobody reads the bounce. It lands in your provider's bounce handling, not in front of a person.
- The address may get suppressed. Many email providers stop sending to an address after a hard bounce. Future emails to that user get dropped too, even after the receiving side is fixed.
2. "Sent" told you nothing
Every one of those messages was "sent" as far as the app knew: the provider accepted the API call. If your dashboard only counts sends, an incident like this is invisible until a support ticket arrives. You need delivered and bounced as separate numbers, plus an alert on the bounce.
3. DMARC is cheap insurance
Microsoft says senders don't need to change anything, and a correctly configured domain can still be hit by someone else's bug. DMARC doesn't make you immune. But during this incident, the domains operators saw refused were the ones without a record, and a p=none record costs nothing, changes nothing about how other receivers treat your mail, and gets you reports about who sends as your domain. Microsoft already requires it above 5,000 messages a day.
The fix: check, publish, align
Check whether you have a record
dig +short TXT _dmarc.example.com
A line starting with "v=DMARC1 means you have one. No output means you don't. Don't trust a dmarc=pass header: Microsoft 365 can stamp dmarc=bestguesspass on mail from domains with no record at all. Ask DNS.
To see why it's empty:
dig TXT _dmarc.example.com +noall +comments | grep status
# status: NXDOMAIN -> the name doesn't exist
# status: NOERROR -> it exists (check the answer section for a TXT record)
Publish a record
Add one TXT record at your DNS host:
| Name | Type | Value |
|---|---|---|
_dmarc |
TXT | v=DMARC1; p=none; rua=mailto:[email protected] |
In zone-file form:
_dmarc.example.com. 3600 IN TXT "v=DMARC1; p=none; rua=mailto:[email protected]"
-
p=nonemeans "report only, take no action on failures". -
rua=is where receivers send daily aggregate reports, which show every source sending as your domain. If the address is on another domain, that domain has to authorise it with a record likeexample.com._report._dmarc.example.net TXT "v=DMARC1". - Publish exactly one. Two DMARC records at the same name is an error.
-
Enter the name as
_dmarcif your DNS panel appends the domain, or you'll create_dmarc.example.com.example.com. -
Already at
p=quarantineorp=reject? Keep it.
Make sure it can pass: alignment
A DMARC record only helps if SPF or DKIM passes for the domain in your From address. In plain terms:
- Your email has a visible From (
[email protected]) and a hidden envelope sender or Return-Path, which is where bounces go. - SPF checks the envelope sender's domain. It aligns if that domain matches your From domain.
-
DKIM checks the signature's
d=domain. It aligns if that matches your From domain. - Alignment is relaxed by default:
d=mail.example.comaligns withFrom: [email protected]because they share an organisational domain.
DMARC passes if either one is aligned.
Many API email providers use their own domain as the Return-Path, so SPF passes for their domain and never aligns with yours unless you set up a custom Return-Path. That's fine: aligned DKIM is usually the path. Make sure your provider signs with your domain (you published its DKIM records), not its own.
Send yourself a message and read the Authentication-Results header. You want something like:
dkim=pass header.d=example.com; dmarc=pass header.from=example.com
If header.d is your provider's domain, DKIM passes but doesn't align. (The SPF/DKIM/DMARC setup guide goes deeper, including moving from p=none to enforcement.)
What about HELO and PTR?
Authentication isn't the only DNS Spamhaus looks at. In its Sender Guidelines announcement of 21 October 2026, Spamhaus says: "Starting in November 2026, we're gradually rolling out enforcement of basic DNS requirements over several months. Spamhaus will require a valid HELO, a reverse DNS (PTR) record, and a matching A record for every sending IP." Its guidelines say failing this "can result in IPs being listed" on its CSS or XBL zones.
Those are properties of the sending IP, so they belong to whoever runs the mail servers. If you send through an API provider, that's the provider, and Spamhaus's ESP guidelines say ESPs "are responsible for the email sent via their infrastructure". Your part is the domain side, which the same guidelines list as a minimum requirement: publish SPF and DMARC for every sending domain, and sign every message with DKIM, aligned with your From domain. They ask that the DKIM d= domain "matches the domain on the header From", so the simplest setup is a d= that's exactly your From domain.
If you run your own mail server, though, the PTR, A record and HELO are yours. Spamhaus has a HELO/DNS checker for that.
Every domain that sends, and sp=
The ones that bite are the forgotten ones: staging. still emailing testers, support. from your helpdesk, billing. from your invoicing tool, the old domain from before a rename. List every domain in a From address and check them all:
for d in example.com mail.example.com support.example.com billing.example.com staging.example.com; do
printf '%-28s %s\n' "$d" "$(dig +short TXT _dmarc.$d | tr -d '\n')"
done
An empty result on a subdomain isn't automatically a gap. Under DMARC, a subdomain with no record of its own falls back to the record on its organisational domain (_dmarc.example.com). That record's sp= tag sets the policy for subdomains, and if there's no sp=, its p= applies. So v=DMARC1; p=none; sp=none; rua=... at the top covers billing.example.com too.
The flip side: if you later move the parent to p=reject without setting sp=, every subdomain inherits reject, including that forgotten staging sender. Set sp= explicitly when you tighten things.
Alert on the right bounce
550 5.7.1 is one of the most generic rejections there is, so alerting on the code alone is noisy. Match on the text, and break it down by sending domain:
// Wherever you process bounces (provider webhook, MTA log parser, etc.).
// `diagnostic` is the raw SMTP response; the field name depends on your source.
const AUTH_REJECT = /5\.7\.(1|23)\b.*(AS9031519|Auth Records|SPF|Sender Policy Framework)/i;
function onBounce({ fromDomain, recipient, diagnostic }) {
if (diagnostic && AUTH_REJECT.test(diagnostic)) {
alert(`Auth-related rejection for ${fromDomain} -> ${recipient}: ${diagnostic}`);
}
}
If your provider doesn't expose the raw diagnostic, alert on a jump in hard bounces to one receiving domain instead. A sudden spike to hotmail.com alone is a signal, whatever the cause. And when an incident like this ends, resend what bounced and review your suppression list for addresses that bounced during the window.
Prove it works
"I added the record" isn't the same as "users get their emails". Close the loop:
-
Confirm the record is public, not just in your DNS panel:
dig +short TXT _dmarc.example.com @1.1.1.1. - Send real mail to a real Hotmail or Outlook.com inbox. Create a free test account and run your actual signup and reset flows against it, more than once.
-
Read the headers there. View the message source and check for
dmarc=passwith your domain. - Track delivered and bounced separately from sent. "Sent" only means your provider accepted the request. There's a Node walkthrough for wiring up delivery and bounce events in Why "sent" isn't "delivered".
The 10-minute checklist
- [ ]
dig +short TXT _dmarc.<domain>for every domain and subdomain in a From address - [ ] Publish one
v=DMARC1; p=none; rua=mailto:...record where it's missing, and setsp=deliberately - [ ] Confirm DKIM signs with your domain (
header.d=matches From) - [ ] Alert on auth-related 5.7.x bounces per sending domain, not just on volume
- [ ] Count delivered and bounced, not just sent
- [ ] Test real flows against a Hotmail inbox
Disclosure: I work on Sendaro, a transactional email API, where DMARC passes via aligned DKIM. Whichever provider you use, the lesson from November is the same: the bounce you don't see is the one that costs you a user.
Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.