Identity problems are sign-in problems
When identity breaks, everything breaks — users cannot sign in, and every app behind the identity provider is unreachable. Most identity incidents fall into a few categories: a sync error in Azure AD Connect, an expired federation certificate breaking SSO, an MFA enrollment loop, or a Conditional Access policy locking out legitimate users. The diagnostic discipline that resolves these fastest is to read the sign-in logs first — they tell you exactly where in the sign-in flow the failure occurred.
This article covers the identity and SSO problems I see most: Azure AD Connect sync errors, ADFS certificate expiry, Okta MFA enrollment loops, and Conditional Access lockouts. For the implementation architecture, see the companion guides on Entra ID SSO implementation and Okta SSO and MFA implementation.
Problem 1: Azure AD Connect sync errors and orphaned objects
A sync error in Azure AD Connect usually surfaces as a user who exists on-premises but not in Entra ID, or a duplicate (two cloud objects for one on-premises user). The most common causes are a duplicate UPN or proxy address across on-premises objects, a filtering rule that excludes the user, or a sourceAnchor mismatch after a migration.
The diagnostic starting point is the Azure AD Connect Sync Service Manager — the Connectors tab shows objects in error with the specific error reason. Resolve the duplicate attribute or filtering cause, then force a delta sync. Never delete a cloud object to “fix” a sync conflict — that orphans the user and breaks their licences and group membership.
- Use Sync Service Manager to find objects in error and the specific reason
- Resolve duplicate UPN or proxy addresses at the on-premises source
- Force a delta sync after fixing the cause: Start-ADSyncSyncCycle -PolicyType Delta
- Never delete a cloud object to resolve a sync conflict — fix the source
Problem 2: ADFS / federation certificate expiry breaking SSO
The most common SSO outage is an expired token-signing or SSL certificate on the ADFS farm or the Exchange federation endpoints. The symptom is sudden — users who signed in yesterday cannot today, and the error points to a trust or certificate problem. Because ADFS certificates auto-renew by default but the relying-party trust must be updated, the renewal silently succeeds on ADFS and silently breaks Office 365.
The fix is to update the federation metadata so Office 365 picks up the new token-signing certificate: run `Update-MsolFederatedDomain` (MSOnline) or the Graph equivalent, and verify the new certificate appears in the relying-party trust. The prevention is a calendar reminder 30 days before every certificate expiry, with a test of the metadata refresh in advance.
# Check the current token-signing certificate on ADFS
Get-AdfsCertificate -CertificateType Token-Signing |
Select-Object Thumbprint, IsPrimary, NotBefore, NotAfter
# Refresh the Office 365 trust with the new certificate
Connect-MsolService
Update-MsolFederatedDomain -DomainName contoso.com -SupportMultipleDomain
# Verify Office 365 sees the new certificate
Get-MsolDomainFederationSettings -DomainName contoso.com |
Select-Object -ExpandProperty IssuerUriProblem 3: Okta MFA enrollment loops
An Okta MFA enrollment loop — the user is repeatedly prompted to set up a factor they already have — is usually caused by a policy requiring a factor the user has not enrolled, or a factor assignment that did not persist. The user appears enrolled but the policy evaluation cannot find a matching factor.
The diagnostic path is the Okta System Log: find the user’s authentication events and read the policy evaluation. If a policy requires a specific factor (e.g. a security key) the user does not have, either enroll the factor or adjust the policy. For admins, keep a break-glass account outside Okta MFA so an MFA outage does not lock out all administrative access.
| Symptom | Cause | Resolution |
|---|---|---|
| Repeated MFA enrollment prompt | Policy requires a factor the user lacks | Enroll the factor or adjust the policy |
| MFA factor not persisting | Assignment failed silently | Re-assign via the System Log trace |
| Admin locked out by MFA | No break-glass account | Keep one admin outside MFA in a vault |
| SSO works but MFA never challenges | MFA policy not bound to the app | Bind the MFA policy to the sensitive app |
Problem 4: Conditional Access locking out legitimate users
A Conditional Access policy that blocks legitimate users is the most disruptive identity problem — and often the intended behaviour misapplied. The common causes: a policy requires a compliant device for an app the user needs on an unmanaged device, a location policy excludes a legitimate office IP, or a policy was enforced without a report-only phase.
The diagnostic path is the Entra ID sign-in logs: open the failed sign-in, view the Conditional Access tab, and see which policy applied and why. If the block is correct (the device is non-compliant), fix the device; if the policy is wrong, adjust it. Always run a new Conditional Access policy in report-only mode for a week to measure impact before enforcing.
- Use the Entra ID sign-in logs to see which CA policy blocked a user and why
- Run new Conditional Access policies in report-only mode before enforcing
- Exclude break-glass admin accounts from all Conditional Access policies
- Review location and device conditions against real user behaviour
Trade-offs and next steps
Identity design balances security against friction. Phishing-resistant MFA (FIDO2, security keys) for admins maximises security but requires hardware and enrollment; push MFA is easier but phishing-susceptible. A strict Conditional Access baseline (compliant device, trusted location) maximises protection but can lock out users during travel or device issues. The right balance is phishing-resistant MFA for admins, a realistic compliance gate with a grace period for users, and report-only mode for every new policy.
When problems recur, the next step is usually to mature the identity foundation: consolidate on a single authoritative directory, automate provisioning and deactivation with SCIM, and keep break-glass accounts outside MFA in a vault. For the implementation architecture, see the Entra ID SSO and Okta SSO/MFA guides.
- Require phishing-resistant MFA for admins; realistic MFA for users
- Run every new Conditional Access policy in report-only mode first
- Keep break-glass admin accounts outside MFA, stored in a vault
- Automate provisioning and deactivation so offboarding is immediate
Lessons learned
- Read the sign-in logs first — they show exactly where the failure occurred
- Put a calendar reminder 30 days before every federation certificate expiry
- Never delete a cloud object to fix a sync conflict — fix the source
- Keep a break-glass admin outside MFA so an outage does not lock out all access
Written by
Jamshed Alam
Global MSP & IT Infrastructure Consultant with 20 years across Exchange, Microsoft 365, Azure, security and automation. Available for project and contract work worldwide.