FLUX workflow causes system freeze when triggered via API (but works in WebUI)ance on resolving this issue.

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

The workflow should run just like it does in the WebUI: progress updates should be received, and image generation should complete successfully.

### Actual Behavior

The terminal shows a progress bar, but it freezes at 0%.

VRAM usage spikes.

The system eventually freezes or crashes (no image is generated).

### Steps to Reproduce

1. Connect to the ComfyUI API via WebSocket using a unique client_id.

2. Send a /prompt request using the FLUX workflow and the same client_id.

3. Wait for progress updates (expected via WebSocket).

4. Observe that the terminal shows a progress bar stuck at 0% and the system becomes unresponsive.

Note: I also tried removing the client_id and polling /history/{prompt_id} instead of using WebSocket, but the problem persisted.

### Debug Logs

```powershell
I don't have logs. Because always system down.
```

### Other

I'm integrating ComfyUI into a Python backend (FastAPI) and using the API to run workflows. While SDXL workflows and LoRA Loader integrations work perfectly, using flux_workflow.json or flux_workflow_gguf.json via API causes the system to hang.

To clarify, the flux_workflow files I'm using are exact copies of the ones that run successfully on the ComfyUI WebUI.

< Environment >
ComfyUI Version: 0.3.48
Workflow: flux_workflow.json and flux_workflow_gguf.json (copied from working WebUI setup)
Backend: Python + FastAPI

OS: Windows 11
RAM: 64GB
GPU: RTX 5080 Laptop GPU

------------------------------
SDXL workflows run flawlessly via the same backend integration.
LoRA Loader also works as expected.
The issue only occurs with FLUX-based workflows.
WebUI handles the same FLUX workflow with no issue.