EVENTHACKER / EDITORIAL
DE EN

Infrastructure Matters · Cloud and communication · EH 121

Category: Infrastructure

When the cloud is running but nobody gets through

🔊 Listen to this article

Listen to the full article. Playback starts when you press Play.

An outage at Microsoft 365 shows how closely email, calendars and collaboration are connected. An explanatory video by Knotenfunke makes the possible technical mechanisms understandable and provides a reason to consider independent communication channels.

Editorial draft · Version 1.0 · 15 November 2026. This article puts the video linked below into context. Details of the specific incident are based on that source; comparison with Microsoft's complete final report is still pending.

A mailbox as a shared dependency

Opening Outlook, sending an appointment, scheduling a meeting: users see these as different actions. Behind the scenes, they may depend on the same infrastructure. The video “Der Microsoft-365-Ausfall, einfach erklärt: warum Outlook mehr als einen Tag hakte” describes an outage starting on 12 November 2026 that reportedly affected Exchange Online and dependent functions in other Microsoft services.

The video distinguishes individual functions: Teams chats and calls reportedly continued working for the most part, while meeting scheduling and invitations were disrupted. That distinction matters for operations. A reachable service can still lose precisely the function a workflow depends on.

Servers can be running and still fail

The video uses a post office as an analogy: mailboxes are in the storeroom, frontend servers are the counters, and a load balancer directs visitors to available counters. To avoid sending requests to a failed server, the system regularly checks its availability. These tests are called health checks.

The principle helps with individual failures. But when many servers stop responding in time simultaneously, work is spread across an ever smaller group. The remaining servers face more load, respond more slowly and may also be removed from distribution. Under unfavourable conditions, protective mechanisms can therefore amplify overload.

Why a restart does not automatically help

Establishing a secure connection takes computing work. Existing connections can often be reused. If large numbers break, clients try to reconnect. A returning server may then receive a wave of new TLS handshakes before it resumes normal operation.

This feedback loop is the video's central technical explanation. It is not, however, a conclusively confirmed account of the Microsoft incident. The narrator explicitly leaves open how the reported network fault was connected in detail to overload during TLS handshakes. The account explains a plausible mechanism; precise confirmation remains for the final report.

What the source reports – and what remains open

According to the video, Microsoft pointed to a problem from an earlier network capacity expansion that became apparent after a later network change. Microsoft also reportedly extended the retry period for undelivered email from 24 to 48 hours. Both details are reproduced here as statements from the video source.

A longer queue can help deliver messages later. For a time-critical decision, however, a late message is not equivalent to one received in time. Nor does the number of outage reports establish the number of individual people affected or a specific financial loss.

The transcript cannot conclusively answer which changes triggered the incident, when all functions became stable again or which measures should prevent a recurrence. This article claims neither a cyberattack nor an incident caused by AI.

What event teams can learn

At events, operational schedules, supplier arrangements and last-minute changes often run through the same communication platform. A second channel adds independence only if it is actually usable when the first fails. A different screen or app alone says little about that.

A useful starting point is a simple inventory: what information does the team need without access to email and calendars? Who has an up-to-date contact list available offline? How are important changes passed on if email fails for a working day? And who decides when to use the fallback?

For time-critical messages, acknowledgement also matters: has the message arrived, and has someone taken responsibility for handling it? Successful sending alone does not answer those questions. The necessary technical and organisational measures depend on the operation and the importance of the message.

Make dependencies visible

The video's value lies in its accessible explanation of a dependency that is hard to see: a cloud consists of many components whose interaction can also complicate recovery. Planning for your own operation therefore starts with a concrete question: which process will still work tomorrow if email and calendars fail tonight?

Source and editing

Knotenfunke: “Der Microsoft-365-Ausfall, einfach erklärt: warum Outlook mehr als einen Tag hakte”, video duration 18:55 minutes. This independent summary is based on a Whisper transcript generated locally on 15 November 2026. It contains some recognition errors, so no verbatim quotations or event times precise to the second are taken from it.

This draft contains no independent confirmation of the specific incident details. Before public release, the stated causes and recovery status should be checked against Microsoft's final report.

Version 1.0 · 15 November 2026 · Unpublished draft