Microsoft’s service-health record, as extracted for one large tenant, shows three Exchange Online incidents between 31 August and 9 September. None shares a cause with the others.
31 August to 3 September: MO1465074
At 14:56 UTC on 31 August, Microsoft opened an incident covering Exchange Online, Microsoft Graph, Teams, Purview, Defender XDR, OneDrive, SharePoint, Universal Print and the 365 admin centre. The stated root cause was a fault in a core authentication configuration shared across those services.
All Exchange Online connection methods were affected. Users could not sign in to Outlook on the web, could not download attachments by any method, could not reach Exchange Web Services, and saw delays or failures on Outlook for iOS and Android. Core mail flow recovered first. Outlook on the web, mobile and Mac lagged because a follow-up mitigation failed to deploy on what Microsoft called less than one percent of its infrastructure.
Microsoft declared impact remediated at 10:00 UTC on 3 September, sixty-seven hours in. A post-incident report was published on 5 September and updated on 8 September. It is visible to tenant administrators only.
4 September: EX1467029
The next day Exchange Online began deferring external mail with 451 4.7.500 Server busy and subcode S77714. Microsoft blamed an anti-spam model, then a throttling rule found still firing during recovery. emailexpert reported it on 10 September. No post-incident report has appeared in the public extract.
Microsoft documents 451 4.7.500 as IP throttling, returned when a sender changes its pattern with higher volume. Senders read it that way. Thomas Johnson told the Mailop list that Cogent fibre cuts had rerouted his San Diego traffic through Los Angeles and put the deferrals down to the shift. Then One.com, Qboxmail, Mailjet and Poppulo reported the same codes across every sending IP and region. Johnson corrected himself. The Cogent cuts were real. They had nothing to do with it.
The deferrals backed up at senders. SMTP2GO reported Microsoft-bound mail stuck, then knock-on delays to Gmail. SuperOffice reported delayed login codes and password resets to Microsoft-hosted recipients.
9 September: EX1469649
At 04:30 UTC on 9 September, Outlook mobile and Outlook for Mac users began failing to connect. Microsoft’s cause: an update intended to improve mail flow was driving high CPU on mailbox database infrastructure. Microsoft built a fix, then decided to revert the update instead. At the time of the extract the incident was open with no completion timeline.
What it means for senders
Authentication configuration, anti-spam model, mail-flow update. Each was a change on Microsoft’s side. Only one of them returned an error code that told the sender to look at itself, and that is the one operators should remember. Microsoft’s guidance for 451 4.7.500 says check your sending pattern. On 4 September that guidance was wrong, and nothing in the response said so.
Microsoft has published a report for MO1465074. For EX1467029, where the error itself misled, the public record has nothing.
