#59 · Six days, seven outages: not hacks, changes
Microsoft, the FAA, United, Chase, Apple, Marriott and the NSA all had outages between 12 and 17 Nov 2026. Almost every stated cause was a change. Safe deployment, rollback and blast radius.
Ek line yaad rakho: duniya hack se kam, update se zyada girti hai.
The week (12–17 Nov 2026)
| Date | Who | Stated cause |
|---|---|---|
| 12–14 Nov | Microsoft 365 | A network capacity upgrade after a recent network change |
| 13 Nov | US FAA | Overloaded switches/routers; linked to a newly installed router |
| 16 Nov | United Airlines | Technology outage, "no indication" of a cyber issue |
| 16 Nov | Chase and other US banks | Chase: a software bug affecting hardware systems |
| 17 Nov | Apple | "An internal update" |
| 17 Nov | Marriott | Sign-in technical issue |
| — | NSA | Internal communications degraded during scheduled maintenance |
Officials say the incidents were unrelated, with no evidence of a coordinated cyberattack (compiled by The National News Desk, 17 Nov 2026).
The pattern
Google's SRE team: "roughly 70% of outages are due to changes in a live system" (Tenets of SRE). This week fits: upgrades, a new router, an internal update, a bug, maintenance.
The interview answer: safe deployment + blast radius
- Small changes, staged: canary to a few users or one region, then widen.
- Detect fast: watch error rates and latency per stage; stop automatically.
- Roll back in one step: every change must be reversible; feature flags help.
- Blast radius: split the system into cells or compartments so one bad change only sinks one compartment.
Words to say: change management, canary, progressive rollout, rollback, feature flags, blast radius, cell-based architecture.
Follow-up questions
If changes cause most outages, why not change less? Freezing changes builds up risk: the next release is bigger and scarier. The goal is many small, safe, reversible changes.
What would you add first to a team with no safe-deployment process? Automatic rollback on error-rate spikes, then staged rollouts by region, then cell isolation for the most critical service.
Causes are the companies' own statements; several root causes were still under investigation.
Go deeper
Part of the Reels track (Notes for every @masstdev reel).
Finished reading?
Mark it done to track your progress in Reels.
#60 · 212 GB at midnight: how GTA 6 unlocked for everyone (19 Nov 2026)
GTA 6 became playable at midnight local time in every country. Why the network didn't have to push 212 GB per player at 12:00, and the thundering-herd answer interviewers want.
#57 · Why did your iPhone ask for the password? The Apple outage (17 Nov 2026)
iCloud sign-in, App Store and Apple Music went down together, and Apple blamed an internal update. Why one login service can take everything down, and the staged-rollout answer interviewers want.