Exchange Online Outage Resilience After EX1550543: Monitoring, Hybrid Continuity and Communication Runbook

On 12 November 2026 at 16:20 UTC, Exchange Online began a global outage that lasted 23 hours 50 minutes (Exchange incident EX1550543, tracked alongside Microsoft 365 incident MO1550821). Outlook, OWA, mobile mail and parts of Teams, SharePoint, OneDrive and Copilot were affected.

You can't prevent a Microsoft-side outage. You can find out before your helpdesk does, keep the business informed, and use the parts of your environment you control. This runbook covers all three.


1. What happened (preliminary PIR)

Microsoft's preliminary post-incident report, as reported publicly, points to three factors that combined:

  1. A latent defect (introduced around July 2026) in a fix for Microsoft 365 networking gateway servers caused lock contention under certain load. CPU and memory were exhausted until the gateways stopped responding. It showed up mainly on high-core-count hardware, which is why lab testing missed it.
  2. A traffic shift. About 15 hours earlier, Teams had failed over to the primary gateway that Exchange Online also uses. That raised load and changed traffic patterns.
  3. Reduced capacity. About 14% of the gateway fleet was out of service for maintenance.

As gateways failed health checks, the remaining servers took more load and also failed. Client connections (MAPI, REST, EAS and OWA) couldn't be established. Microsoft ruled out a cyberattack. Recovery involved pausing deployments, rolling back configuration and bringing restarted servers back gradually.

The lesson for customers: the failure was in shared front-door infrastructure, so no tenant setting could have avoided it. What you control is detection, communication and fallback.


Outage response timeline 0-15 min: Detect Service health poll Synthetic mail probe Helpdesk spike alert 15-60 min: Confirm Match incident ID Rule out local cause First user notice 1 h+: Operate Cached mode guidance Watch hybrid queues Hourly status updates After: Recover Queues drained Message trace checks Read the final PIR The goal is to know before your users tell you, and to tell them before they ask.

2. Detect: poll the Graph service health API

The Microsoft 365 admin center can be slow to load during a big incident. Poll the Graph service communications API from automation instead. It needs the ServiceHealth.Read.All application permission.

POWERSHELL
# App-only connection (certificate-based)
Connect-MgGraph -ClientId $AppId -TenantId $TenantId -CertificateThumbprint $Thumbprint -NoWelcome

# Current health of mail-related services
Get-MgServiceAnnouncementHealthOverview |
    Where-Object { $_.Service -match 'Exchange|Teams|SharePoint|Microsoft 365 suite' } |
    Select-Object Service, Status |
    Format-Table -AutoSize

# Open incidents affecting Exchange Online
Get-MgServiceAnnouncementIssue -Filter "service eq 'Exchange Online' and isResolved eq false" |
    Select-Object Id, Title, Classification, Status, StartDateTime, LastModifiedDateTime |
    Format-Table -AutoSize

Lab result: Run against the LAB 365 dev tenant without ServiceHealth.Read.All, the API returned HTTP 403. Directory permissions don't cover it, so add this permission explicitly to your monitoring app.

Turn it into an alert

Schedule this every 5 minutes (Azure Automation or Task Scheduler, see our unattended reporting guide) and post to Teams or email when an incident appears:

POWERSHELL
$issues = Get-MgServiceAnnouncementIssue -Filter "service eq 'Exchange Online' and isResolved eq false and classification eq 'incident'"

foreach ($i in $issues) {
    $stateFile = "C:\Monitoring\seen-$($i.Id).txt"
    if (-not (Test-Path $stateFile)) {
        $body = @{ text = "Exchange Online incident $($i.Id): $($i.Title) (status: $($i.Status))" } | ConvertTo-Json
        Invoke-RestMethod -Method Post -Uri $TeamsWorkflowUrl -Body $body -ContentType 'application/json'
        New-Item -Path $stateFile -ItemType File | Out-Null
    }
}
TIP:
If Teams is down too, your Teams alert won't arrive. Send the alert to a second channel as well, such as SMS through your paging tool or an external email address.

3. Detect: synthetic mail probe

Service health can lag behind real impact. A simple probe sends a message every 15 minutes from an external mailbox to a monitored internal mailbox and alerts if it doesn't arrive within 10 minutes. Most monitoring tools (PRTG, SolarWinds, Azure Monitor with Logic Apps) have a ready-made mail round-trip sensor. Use it.


4. Operate: what hybrid organizations can still do

During EX1550543, the problem was clients reaching Exchange Online. For hybrid organizations:

  • On-premises mailboxes kept working. Users on Exchange Server SE could send and receive. Mail to cloud mailboxes queued on the on-premises side instead of being lost.
  • Watch your hybrid send connector queue so you know the backlog and can confirm it drains afterwards:
POWERSHELL
# Exchange Server SE / 2019 Management Shell
Get-Queue | Where-Object { $_.NextHopDomain -like "*mail.onmicrosoft.com" } |
    Select-Object Identity, Status, MessageCount, LastError |
    Format-Table -AutoSize

# Summary across all transport servers
Get-QueueDigest -Dag (Get-DatabaseAvailabilityGroup).Name -GroupBy NextHopDomain
  • Check queue expiry. The default MessageExpirationTimeout is 2 days, so a 24-hour outage fits within it. Confirm no one has lowered it:
POWERSHELL
Get-TransportService | Select-Object Name, MessageExpirationTimeout
  • Tell users about Outlook cached mode. Users in cached mode can still read and search mail already downloaded, and new messages wait in the Outbox. Don't advise users to recreate profiles during an outage. It deletes their local cache and leaves them with nothing.

5. Recover: confirm nothing was lost

After Microsoft declares the incident resolved:

POWERSHELL
# On-premises: hybrid queue should return to zero
Get-Queue | Where-Object { $_.NextHopDomain -like "*mail.onmicrosoft.com" } | Select-Object Identity, MessageCount

# Exchange Online: spot-check delivery for the outage window
Connect-ExchangeOnline
Get-MessageTraceV2 -StartDate "2026-11-12 16:00" -EndDate "2026-11-13 18:00" -Status Failed |
    Group-Object RecipientAddress | Sort-Object Count -Descending | Select-Object -First 20 Count, Name

Then read Microsoft's final PIR in the admin center (Health → Service health → History), and record the incident ID and business impact for your service review or credit claim.


6. Communication checklist

WhenMessageChannel
Within 30 min of confirmation"Microsoft is having an email outage (ID EX…). Your mail isn't lost. Use cached Outlook. Next update at HH:MM."SMS, intranet banner, phone tree
Every hourStatus, Microsoft's latest update, next update timeSame channels
Resolution"Service is restored. Delayed messages are arriving. Report anything still missing to the helpdesk."Email plus intranet
Within 5 daysSummary with root cause (from the PIR) and actions takenEmail to stakeholders
Prepare these templates now, and store them somewhere that doesn't depend on Microsoft 365.