Entra ID access review stuck in initializing state: timeout troubleshooting

TL;DR

Wait a measured interval and check the instance status through Graph before you delete anything, because large scopes legitimately take time to build. Note the scope, the number of principals and whether reviewers are users, managers or group owners, since reviewer resolution is the step that most often stalls. If the status never moves, delete the stuck instance, fix the scope or reviewer source that stalled it, and let the schedule create a fresh instance rather than forcing the broken one.

Entra ID access review stuck in initializing state: timeout troubleshooting

Use this when

• A new access review instance stays at initializing long past the usual start window.
• Reviewers never receive tasks although the schedule fired on time.
• Only reviews with very large groups or manager reviewers stall.

Not for this skill when

• The review is in progress and reviewers are deciding. That is normal operation, not a stall.
• The definition itself is missing. Work the definition lookup problem instead.

Steps

1. Record when the instance was created, its scope and its reviewer type, and check the Microsoft service health view for identity governance incidents in the same window. Verify: creation time, scope and reviewer type are noted and service incidents are ruled in or out.
2. Query the instance status through Microsoft Graph after at least one normal initialization window and compare it with a smaller review that started successfully in the same tenant. Verify: both statuses are recorded so tenant wide and instance specific causes separate cleanly.
3. Count the principals in the review scope and check for nested groups or dynamic groups whose membership calculation is itself slow or failing. Verify: the scope size is known and any problem group is named.
4. If reviewers are managers, sample several in-scope users and confirm each has a manager value in Entra ID, since mass missing managers can stall reviewer resolution. Verify: sampled users show manager values or the missing set is listed.
5. Delete the stuck instance only, leaving the schedule definition intact, then fix the named cause such as trimming the scope or switching reviewer type for this cycle. Verify: the stuck instance is gone and the definition still exists with its schedule.
6. Start a replacement instance or wait for the next scheduled one, watch it move to in progress, and confirm reviewers can open items. Verify: the new instance reaches in progress and a reviewer opens an item.
7. Document the stall, its cause and the working reviewer configuration in the cycle evidence so the next schedule reuses the fixed setup. Verify: the evidence names the cause rather than only noting a delay.

Why it happens

Initialization is when Entra expands the scope into individual review items and resolves every reviewer. Huge groups, dynamic membership that recalculates slowly, and manager based review across a directory with patchy manager data all stretch that phase, and a failure inside it leaves the instance parked with little surface detail. Comparing with a smaller successful review in the same tenant isolates whether the tenant, the scope or the reviewer source is at fault, which decides whether the fix is patience, scope design or directory data.

Variants and edge cases

Entra access review not starting for reviewers

The reviewer phrasing. Reviewers see nothing because no instance finished building; check instance status before resending invitations.

access review instance stuck notstarted or initializing in Graph

The Graph phrasing. Poll status at intervals; a status that changes, even slowly, is a big review, not a stuck one.

large group access review times out on creation

The scale phrasing. Split the group into department sized reviews that initialize reliably and report together.

Edge cases to check before closing the review cycle:

• A review scoped to an application with many service principals builds slowly. Give service principal heavy scopes extra time before intervening.
• Deleting the definition instead of the instance destroys history. Always delete at instance level for this fault.
• Tenant migrations can strand instances. Check for a recent directory change if nothing else explains the stall.

Compatibility: Microsoft Entra ID Identity Governance access reviews, Microsoft Graph access review instance endpoints, Entra admin center.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_hLvAuLPnLAaBY1pW8JVJYg