Content Understanding calls failing with "InternalServerError - An unexpected error occurred"

Since the afternoon of June 9 (US Eastern), Content Understanding analyze calls that previously worked reliably have been failing on our resource in East US.

Setup: custom extract analyzer (gpt-4.1, created January 2026), called via POST /contentunderstanding/analyzers/{analyzerId}:analyzeBinary?api-version=2025-11-01.

Symptom: the submit succeeds (202 with a valid Operation-Location), but the operation transitions to Failed roughly four seconds later with:

{"code":"InternalServerError","message":"An unexpected error occurred."}

No innerError, empty contents. This happens for multiple files across separate environments.

What I've ruled out:

• It is not our client code: I can reproduce the identical error body in Content Understanding Studio itself, testing against the same project. Studio uses the project's own draft analyzer, so it is also not specific to our published analyzer.
• It is not file content: the same unmodified files pass on some runs and fail on others. The failures are intermittent in Studio and near-constant from our application.
• It is not auth or request format: submissions are accepted with 202; the failure occurs during backend analysis.

Trace details:

• Studio failure: request-id / apim-request-id 824b4f37-f63b-4aa8-9636-0f3e908c06fa, created 2026-06-10T14:25:40Z, x-ms-region: East US
• Application failure: operation 077b4af9-3a8e-4e0e-b522-1d4622c9e4d0, submitted 2026-06-10T13:46Z

Can someone trace these request IDs and confirm whether a recent rollout in East US is affecting the analyze path for custom analyzers on api-version 2025-11-01? Happy to provide additional failing operation IDs — this is blocking a production workload.