QWEN EDIT 2511 AIO model image generation fails to work properly after version 0.22.

### 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

This issue has persisted from version 0.23 through the latest version, 0.27. When generating images using the QWEN EDIT 2511 AIO model, VRAM usage was around 80% prior to version 0.23; however, with versions 0.23 and later, the VRAM fills up completely and the system spills over into using system memory. Image generation becomes extremely slow—effectively rendering the tool unusable.

### Actual Behavior

RT

### Steps to Reproduce

RT

### Debug Logs

```powershell
RT
```

### Other

_No response_

Start by reproducing QWEN EDIT 2511 AIO image generation on version 0.22 and on versions 0.23 through 0.27, with custom nodes disabled, while recording VRAM and system-memory usage. No source files or tests are named in the report, so trace the model-loading and generation path after confirming the regression. Done means generation no longer spills into system memory or otherwise returns to the pre-0.23 resource behavior.

Written by the indexing model from the issue text.