Microsoft finally explained its 20-hour mail outage, and the explanation should worry you more than the outage did. Gmail's inbox placement numbers for 2026 are in, and they're ugly. And France's top court just told advertisers that renting space inside someone's inbox doesn't get you out of consent law. Let's get into it.
Microsoft has published its explanation for the Nov 12–13 outage that took down Outlook.com, Hotmail, Exchange Online, and Microsoft 365 mail worldwide for most of a business day. Downdetector logged over 97,000 reports at peak. About 80% of the networking-gateway servers behind Microsoft 365 mail went unresponsive.
The root cause: a latent software defect, introduced by a July 2026 fix to the Microsoft 365 networking gateway, that sat undetected for months because Microsoft's load tests ran on different hardware and traffic patterns than production actually sees. It finally surfaced when a Teams gateway failover roughly 15 hours earlier dumped extra load onto the affected gateway — one that was already running at reduced capacity, since about 14% of it was offline from an unrelated issue. Microsoft's final post-incident report is due by Nov 20.
This isn't a story about Microsoft having a bad day. It's a story about a bug that shipped in July, survived every subsequent load test, and only announced itself four months later under a specific, unlucky combination of failover traffic and partial capacity loss. If your load-testing environment doesn't match production hardware and traffic shape, you don't actually know what you're protected against — you know what you tested against. That gap is exactly where this bug lived for four months, and it's the second distinct Microsoft mail-reliability incident in consecutive issues of this newsletter.
Sources: BleepingComputer, The Register, SecurityWeek, CNN
Sponsored

Battle-Tested Email Infrastructure Used By the World's Largest Senders
When you choose GreenArrow, you aren’t just buying another product or service. You’re choosing an experienced partner who understands the industry and your place in it. We provide proven, battle-tested tools along with the expert guidance to implement them effectively.
Whether you need maximum control with GreenArrow Engine on-prem or a turnkey managed solution with GreenArrow CloudMTA, we’ve got a modern deployment option that fits your use case.
If you’re bootstrapping an email program from the ground up, undergoing rapid scaling, or migrating a mature operation from another vendor, we’ve done it before with customers just like you.
Learn More and Get a Free DemoHere's the number that should get your attention: personal Gmail inbox placement fell from 90.97% in Q1 2026 to 81.57% in Q3, according to Validity's Email Radar data. Google Workspace placement fell even harder, from 88.9% down to 74.3% over the same stretch. Spam-foldering on personal Gmail rose from 5.4% to 9.5%.
Validity points at a cluster of candidate causes — Promotions-tab changes, Gmail's AI-driven inbox-summary features, and tighter bulk-sender enforcement — but is upfront that its data can't isolate which one is actually driving the decline. That's an honest caveat, and it's also the catch: you can't fix what you can't attribute.
Nine points of inbox placement isn't noise, and a drop this size heading into Black Friday and Cyber Monday is the kind of thing that turns into a Monday morning crisis call if you're not already watching it. Check your own Gmail-specific placement against this baseline now, not after your biggest sends of the year go out.
Sources: Validity, "Email Radar: The Great Gmail Inbox Placement Crash", emailexpert
France's Conseil d'État — the country's highest administrative court — has upheld a €50M fine against Orange for selling ad space inside Mail Orange inboxes, in a ruling from Oct 16 that surfaced in English-language coverage this week. The ads were formatted to look like regular emails and were seen by more than 7.8 million people.
Orange argued that mailbox-level ad placement falls outside direct-marketing rules because it never actually touched a recipient's email address. The court rejected that argument outright, ruling the placements count as electronic direct marketing requiring specific prior consent. It also upheld a second finding: Orange kept reading tracking cookies after users had withdrawn consent.
The ruling matters well beyond Orange. It establishes that buying ad space from a mailbox operator doesn't get you a free pass on consent requirements just because you never saw, rented, or sent to the actual address. If you're a platform or advertiser eyeing inbox-ad placements as a consent-law workaround, this is the precedent that closes that door — with a daily penalty attached for noncompliance.
Sources: Next.ink, UniversFreebox, emailexpert
Klaviyo's Q3 revenue grew 25% year-over-year to $387.2M. Its $50K+ ARR customer cohort grew 33% to 4,725 accounts, and the bigger cohorts grew faster still — 49% for $250K+ accounts, 57% for $500K+. The board also approved a $300M buyback.
The numbers are the easy part. The repositioning is the story: Klaviyo is now explicitly marketing itself as an "autonomous B2C CRM" — a customer data platform plus AI agents making send, segment, and channel decisions across email, SMS, RCS, WhatsApp, and push, rather than just executing campaigns a human marketer built.
Every major ESP is racing toward some version of "the AI decides what goes out." The growth numbers say customers are buying it. The open question is how that lands with brands that still want a human eyeballing what ships under their name before it ships — and whether "autonomous" survives contact with the first campaign a bot sends that nobody signed off on.
Sources: emailexpert, Klaviyo Q3 2026 results announcement
Disclosure: I work at Klaviyo. This item covers only their public Q3 earnings release and outside reporting on it — nothing here is non-public.
Between Nov 7 and Nov 14, an attacker created roughly 570 free-trial accounts on course platform LearnWorlds, bypassed the per-customer account limits, and abused the platform's automated welcome-email feature to blast out 13,000 fake YouTube copyright-infringement notices. The payload was OAuth consent-phishing designed to hijack victims' Google and YouTube accounts.
LearnWorlds says its core systems weren't breached — this was abuse of a legitimate feature, not an intrusion. The company has since added bot detection and tighter trial fraud-scoring.
This is the pattern worth internalizing: the attacker didn't build phishing infrastructure. They rented it, one free trial at a time, from a SaaS platform's own transactional-email pipeline — the welcome emails every signup flow sends automatically, with a sender reputation already warmed up for them. If you run a platform with free-trial signup and automated transactional mail, this is a live abuse vector, not a one-off incident to read about and move on from.
Source: The Register
app.from_email is now an array, not a string. GitHubThat's the week. If something is wrong, reply and tell me — I read every response.
— John
This Week In Email — thisweekin.email
Subscribe to This Week in Email and get future issues delivered to your inbox.