Hugging Face Spaces stuck in APP_STARTING due to scheduling failure

Users are reporting that Hugging Face Spaces, particularly on the free CPU-basic tier, remain indefinitely stuck in the `APP_STARTING` stage without allocating hardware. The API reports `hardware.current: null` while the UI shows a spinner, and public endpoints time out with HTTP 000.

• The issue appears to be a scheduling failure where no machine is allocated despite requests for `cpu-basic` hardware.
• Restarting or performing a factory rebuild does not resolve the problem, as containers immediately return to the starting state.
• Logs eventually indicate a "Scheduling failure: unable to schedule" error after the initial clean startup logs.

This suggests a capacity shortage on Hugging Face's side rather than an application-level bug, leaving users with no option but to wait for resolution.