Power BI Embedded capacity in zombie/inconsistent ARM state — disappears from Portal, ResourceNotFou

Hi,

I have a Power BI Embedded capacity (pbiembeddedeus2) that repeatedly enters an inconsistent/zombie state. I need help to either force-delete it (so the name is released and I can recreate it) or restore it to a healthy Succeeded state, and to understand the root cause, since the problem keeps recurring.

Environment:

• Subscription: WKS PRODUCTION - PT (0a75ee25-18c2-4628-b3b2-5f534f325511)
• Resource Group: RG-POWERBI-EUS2
• Resource: pbiembeddedeus2
• Type: Microsoft.PowerBIDedicated/capacities
• Region: East US 2

Symptoms (recurring over multiple days):

1. The capacity intermittently disappears from the Portal — the Resource Group shows as empty, no resources listed.
2. ARM API returns ResourceNotFound (404) when querying or trying to resume the capacity.
3. Portal blocks creation of a new resource with the same name: "A Power BI resource with the same name already exists in the selected location" — so the name is held, but the resource is unreachable (classic zombie/control-plane inconsistency).
4. az resource list returns empty across all subscriptions I have access to.
5. No delete operation for the capacity or the resource group appears anywhere in the Activity Log (checked the full 90-day window, scoped directly to the RG). The resource is also not present in any soft-delete list.
6. Resource Health shows repeated "Unavailable (Customer Initiated)" events on three consecutive days (2026-07-29, 07-30, and 07-31), which lines up with the recurring nature of the issue.

Recurrence:

• 2026-07-30: capacity disappeared from the Portal after a resume attempt; ARM returned ResourceNotFound.
• 2026-07-31 (today): the resource reappeared in the Portal but paused; attempting to start it triggered the exact same symptom again — it disappeared once more.

Correlation IDs for investigation:

• Failed Resume operation: a669a1de-887c-4faa-96c7-9f63be4962af
• Portal snapshot error: 38fe582f-a0cd-4f8f-8c2c-5c6a81d852fd
• Role assignment delete event: 08927383-e6c6-4a47-8046-0f409beea027

Timeline:

• 2026-07-29 ~23:16 UTC: automationweur SP suspended the capacity (routine daily schedule — worked fine for the prior weeks).
• 2026-07-30 ~15:22 UTC: A Contributor role assignment scoped to the capacity, assigned to the automationweur SP (principalId fab5998f-623f-4036-a274-2839d9249a1b), was removed. This event has no caller and no claims, which suggests an automatic platform cleanup rather than a user action.
• 2026-07-30 ~15:23 UTC: Manual Resume attempted — failed. Resource entered the inconsistent state.
• 2026-07-31: Resource briefly reappeared, then disappeared again on start attempt (same symptom).

What I have already ruled out:

• Wrong subscription/tenant context (confirmed correct via az account show; checked all accessible subscriptions).
• Wrong Portal filters (All resources with all filters cleared).
• No active Service Issue is reported against my subscription (Resource Health events API returns empty for eventType ServiceIssue), so this does not appear to be a broad regional outage.

Request: Please investigate the suspected ARM control-plane / metadata inconsistency for this capacity. Specifically:

1. Force-delete the orphaned resource entry so the name is released and I can recreate it, OR restore it to a healthy Succeeded state.
2. Since the Activity Log has no delete event, please check the internal Microsoft.PowerBIDedicated resource provider logs to identify what is driving the resource to disappear — as this has now recurred across three days, I need the root cause, not just a one-off fix.

Thank you.

Note: I've also posted this on Microsoft Q&A for visibility: https://learn.microsoft.com/en-us/answers/questions/5962621/power-bi-embedded-capacity-in-zombie-inconsistent