ZeroGPU Space stuck at APP_STARTING with “Scheduling failure: unable to schedule”

Hi Hugging Face team,

Our ZeroGPU Space is failing to start:

The build completed successfully, but the Space remained in the starting state and eventually reported:

Scheduling failure: unable to schedule

We tried a normal restart, followed by a pause/resume cycle. Neither restored the app during our observation window. As of October 27, 2026, 06:30 UTC, the runtime API returned:

{
  "stage": "APP_STARTING",
  "hardware": {
    "current": null,
    "requested": "zero-a10g"
  },
  "replicas": {
    "requested": 1
  }
}

The latest container log contains only:

===== Application Startup at 2026-10-27 06:27:24 =====

There is no Python traceback or application-ready message.

The Space was previously working. The latest commit (72851f1) only adds header links and minor CSS changes; the dependencies, model-loading code, and hardware configuration are unchanged.

Could you please check whether this Space is affected by a scheduling or resource-allocation issue, and help restore it?

Thank you!