Python crashes with no logs when trying to use Qwen-Image-Edit

### 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

I expect Qwen Image Edit workflow to work.
I expect ComfyUI Python to not crash.
I expect errors to be reported in the log.

### Actual Behavior

Some time after the last log line ("loaded completely; 21479.77 MB usable, 7910.29 MB loaded, full load: True"), Python crashes with a non-descriptive error window. Nothing is printed in the logs. Lots of free RAM and VRAM.

### Steps to Reproduce

1. Be a new user of ComfyUI, so the system has no old leftovers or user files.
2. Download and unpack a fresh clean version of ComfyUI. I've tested:

* ComfyUI portable v0.8.0
* ComfyUI portable v0.6.0
* ComfyUI Desktop

2. Go to Workflow Templates, search for "Qwen" and open the "Imaging Edit (new)" workflow template (02_qwen_Image_edit_subgraphed): https://docs.comfy.org/tutorials/image/qwen/qwen-image-edit (or the "image_qwen_Image_2512" template)
3. Select input image and start generation.
4. Some time after the last log line ("loaded completely; 21479.77 MB usable, 7910.29 MB loaded, full load: True"), Python crashes with a non-descriptive error window. Nothing is printed in the logs. Lots of free RAM and VRAM.

### Debug Logs

```powershell
Checkpoint files will always be loaded safely.
Total VRAM 24575 MB, total RAM 130997 MB
pytorch version: 2.9.1+cu130
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 58948.0
working around nvidia conv3d memory bug.
Using pytorch attention
Python version: 3.13.9 (tags/v3.13.9:8183fa5, Oct 14 2025, 14:09:13) [MSC v.1944 64 bit (AMD64)]
ComfyUI version: 0.6.0
****** User settings have been changed to be stored on the server instead of browser storage. ******
****** For multi-user setups add the --multi-user CLI argument to enable multiple user profiles. ******
ComfyUI frontend version: 1.34.9
[Prompt Server] web root: ComfyUI_windows_portable_nvidia_v0.6.0\python_embeded\Lib\site-packages\comfyui_frontend_package\static
Total VRAM 24575 MB, total RAM 130997 MB
pytorch version: 2.9.1+cu130
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 58948.0

Import times for custom nodes:
0.0 seconds: ComfyUI_windows_portable_nvidia_v0.6.0\ComfyUI\custom_nodes\websocket_image_save.py

Context impl SQLiteImpl.
Will assume non-transactional DDL.
No target revision found.
Starting server

To see the GUI go to: http://127.0.0.1:8188
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
Requested to load WanVAE
loaded completely; 19624.97 MB usable, 242.03 MB loaded, full load: True
Found quantization metadata version 1
Using MixedPrecisionOps for text encoder
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load QwenImageTEModel_
loaded completely; 21479.77 MB usable, 7910.29 MB loaded, full load: True

```
main.log (For ComfyUI desktop)
```
...
[2026-01-11 01:44:05.922] [info] @todesktop/runtime: Update for version 0.7.2 is not available (latest version: 0.7.2, downgrade is allowed).
[2026-01-11 01:44:05.923] [info] @todesktop/runtime: AutoUpdater: update-not-available {"version":"0.7.2","files":[{"url":"ComfyUI Setup 0.7.2 - Build 26010815jy3gb43-x64.exe","sha512":"E3Fx90aOOU72KsXqsZwOi2aEsu3dIZkC1PpSwRzmyg426ynHjPN50Sse20sEeqehO1OSLYOfrrzvV4sdg4j8eA==","size":133856824}],"path":"ComfyUI Setup 0.7.2 - Build 26010815jy3gb43-x64.exe","sha512":"E3Fx90aOOU72KsXqsZwOi2aEsu3dIZkC1PpSwRzmyg426ynHjPN50Sse20sEeqehO1OSLYOfrrzvV4sdg4j8eA==","releaseDate":"2026-01-08T13:41:58.873Z"}
[2026-01-11 01:44:49.073] [info] Python server is ready
[2026-01-11 01:49:05.163] [error] Python process exited with code 3221225477 and signal null
```

### Other

System: Windows 10 Pro latest
GPU: NVidia RTX 3090 24GB
RAM: 128GB
CPU: AMD Ryzen 5950 (32 cores)

SD XL generation works.
Latest NVidia studio drivers: 591.44 / 32.0.15.9144

Windows event log:

```
Faulting application name: python.exe, version: 3.13.9150.1013, time stamp: 0x68ee5a66
Faulting module name: torch_cpu.dll, version: 0.0.0.0, time stamp: 0x690e9551
Exception code: 0xc0000005
Fault offset: 0x00000000066187b9
Faulting process id: 0x249c8
Faulting application start time: 0x01dc82d950a1cd1c
Faulting application path: ComfyUI_windows_portable_nvidia_v0.6.0\python_embeded\python.exe
Faulting module path: ComfyUI_windows_portable_nvidia_v0.6.0\python_embeded\Lib\site-packages\torch\lib\torch_cpu.dll
Report Id: 6a427b5f-93de-47fe-a5ba-09ac0c62346c
Faulting package full name:
Faulting package-relative application ID:
```

Tested Torch and CUDA and it works (used the ComfyUI Portable 0.8.0 embedded Python):
```
>>> torch.rand(50000, 30000).cuda() @ torch.rand(30000, 20000).cuda()
tensor([[7458.6982, 7410.8911, 7395.3438, ..., 7485.6777, 7477.3555,
7494.6875],
[7477.8164, 7453.9629, 7443.1455, ..., 7533.3398, 7516.6699,
7517.3984],
[7503.0737, 7512.5317, 7505.6343, ..., 7595.6221, 7552.3735,
7583.5098],
...,
[7459.0713, 7442.6699, 7441.1719, ..., 7460.5400, 7486.6460,
7504.0264],
[7489.3799, 7447.6631, 7440.9907, ..., 7508.7471, 7502.6880,
7550.0713],
[7490.8721, 7478.8467, 7445.2734, ..., 7537.7051, 7527.1514,
7526.8086]], device='cuda:0')
```