LTX 2.3 repeated-run system freeze on RTX 3080 Ti resolved by disabling comfy-kitchen CUDA backend

### 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

ComfyUI should be able to process a queued LTX 2.3 video workflow repeatedly without freezing the host system.

If a generation fails due to VRAM pressure, kernel failure, or backend incompatibility, I would expect ComfyUI/PyTorch to raise an error, cancel the prompt, or recover after the prompt, not hard-freeze the whole desktop/OS.

The same queue should continue to run with the desktop remaining responsive.

### Actual Behavior

With ComfyUI v0.26.0 and `comfy-kitchen` CUDA backend enabled, repeated LTX 2.3 video generations freeze the whole system after approximately 3 runs. The machine becomes unresponsive rather than ComfyUI throwing a recoverable error.

This reproduced with more than one LTX 2.3 workflow, and also reproduced on CachyOS with NVIDIA drivers. I can provide the exact CachyOS NVIDIA driver version from `nvidia-smi`.

On Windows, the failing log shows `comfy-kitchen` CUDA enabled and then repeated:

`Model LTXAV prepared for dynamic VRAM loading...`

before the log stops.

When I locally disabled only the `comfy-kitchen` CUDA backend, the same kind of LTX 2.3 queue progressed much further: at least video 8 of 12, with the desktop still responsive, and videos completing in under 600 seconds.

### Steps to Reproduce

Setup:
Windows 11 and CachyOS on same machine
RTX 3080 Ti, 32Gb VRAM
ComfyUI v0.26.0/v0.27.0
torch 2.10.0+cu130 / CUDA 13.0
comfy-kitchen CUDA backend enabled
Any LTX 2.3 Workflow (15s 24fps 1024x1024 2-3 passes)

Limited set of custom nodes:
.\python_embeded\python.exe -I -W ignore::FutureWarning ComfyUI\main.py --windows-standalone-build --lowvram --reserve-vram 2 --disable-pinned-memory --disable-all-custom-nodes --whitelist-custom-nodes LTX2EasyPrompt-LD ComfyUI-LTXVideo comfyui-kjnodes ComfyUI-GGUF comfyui-videohelpersuite ComfyMath ComfyUI-Impact-Pack nevilex-utils --port %PORT%

1. Install/run ComfyUI v0.26.0 with the default `comfy-kitchen==0.2.10` dependency.
2. Use an NVIDIA GPU system. My Windows test system uses an RTX 3080 Ti. The issue also reproduced on CachyOS with NVIDIA drivers.
3. Run an LTX 2.3 video workflow that loads `LTXAV` and uses dynamic VRAM/model loading.
4. Queue multiple video generations, for example 7 runs.
5. Leave `comfy-kitchen` CUDA backend enabled.
6. Observe that after approximately 0-3 completed/started runs, the whole system freezes.
7. As a comparison test, disable only the `comfy-kitchen` CUDA backend and rerun the same queue.
8. Observe that the queue progresses much further, at least 8 videos in my test, while the desktop remains responsive.

### Debug Logs

```powershell
[2026-07-01 19:29:33.164] [LTX2-GGUF] Flushing ComfyUI/Torch memory after prompt worker...
[2026-07-01 19:29:33.578] [LTX2-GGUF] ComfyUI soft cache flush complete.
[2026-07-01 19:29:33.580] [LTX2-GGUF] VRAM after flush: 0.01GB allocated / 0.03GB reserved
[2026-07-01 19:29:34.465] [LTX2-GGUF] FPS output: 24 (preset: None — let the LLM decide, portrait: False)
[2026-07-01 19:29:34.592] Found quantization metadata version 1
[2026-07-01 19:29:35.476] clip missing: ['multi_modal_projector.mm_input_projection_weight', 'multi_modal_projector.mm_soft_emb_norm.weight', 'vision_model.embeddings.patch_embedding.weight', 'vision_model.embeddings.patch_embedding.bias', 'vision_model.embeddings.position_embedding.weight', 'vision_model.encoder.layers.0.layer_norm1.weight', 'vision_model.encoder.layers.0.layer_norm1.bias', 'vision_model.encoder.layers.0.self_attn.q_proj.weight', 'vision_model.encoder.layers.0.self_attn.k_proj.weight', 'vision_model.encoder.layers.0.self_attn.v_proj.weight', 'vision_model.encoder.layers.0.self_attn.out_proj.weight', 'vision_model.encoder.layers.0.layer_norm2.weight', 'vision_model.encoder.layers.0.layer_norm2.bias', 'vision_model.encoder.layers.0.mlp.fc1.weight', 'vision_model.encoder.layers.0.mlp.fc2.weight', 'vision_model.encoder.layers.1.layer_norm1.weight', 'vision_model.encoder.layers.1.layer_norm1.bias', 'vision_model.encoder.layers.1.self_attn.q_proj.weight', 'vision_model.encoder.layers.1.self_attn.k_proj.weight', 'vision_model.encoder.layers.1.self_attn.v_proj.weight', 'vision_model.encoder.layers.1.self_attn.out_proj.weight', 'vision_model.encoder.layers.1.layer_norm2.weight', 'vision_model.encoder.layers.1.layer_norm2.bias', 'vision_model.encoder.layers.1.mlp.fc1.weight', 'vision_model.encoder.layers.1.mlp.fc2.weight', 'vision_model.encoder.layers.2.layer_norm1.weight', 'vision_model.encoder.layers.2.layer_norm1.bias', 'vision_model.encoder.layers.2.self_attn.q_proj.weight', 'vision_model.encoder.layers.2.self_attn.k_proj.weight', 'vision_model.encoder.layers.2.self_attn.v_proj.weight', 'vision_model.encoder.layers.2.self_attn.out_proj.weight', 'vision_model.encoder.layers.2.layer_norm2.weight', 'vision_model.encoder.layers.2.layer_norm2.bias', 'vision_model.encoder.layers.2.mlp.fc1.weight', 'vision_model.encoder.layers.2.mlp.fc2.weight', 'vision_model.encoder.layers.3.layer_norm1.weight', 'vision_model.encoder.layers.3.layer_norm1.bias', 'vision_model.encoder.layers.3.self_attn.q_proj.weight', 'vision_model.encoder.layers.3.self_attn.k_proj.weight', 'vision_model.encoder.layers.3.self_attn.v_proj.weight', 'vision_model.encoder.layers.3.self_attn.out_proj.weight', 'vision_model.encoder.layers.3.layer_norm2.weight', 'vision_model.encoder.layers.3.layer_norm2.bias', 'vision_model.encoder.layers.3.mlp.fc1.weight', 'vision_model.encoder.layers.3.mlp.fc2.weight', 'vision_model.encoder.layers.4.layer_norm1.weight', 'vision_model.encoder.layers.4.layer_norm1.bias', 'vision_model.encoder.layers.4.self_attn.q_proj.weight', 'vision_model.encoder.layers.4.self_attn.k_proj.weight', 'vision_model.encoder.layers.4.self_attn.v_proj.weight', 'vision_model.encoder.layers.4.self_attn.out_proj.weight', 'vision_model.encoder.layers.4.layer_norm2.weight', 'vision_model.encoder.layers.4.layer_norm2.bias', 'vision_model.encoder.layers.4.mlp.fc1.weight', 'vision_model.encoder.layers.4.mlp.fc2.weight', 'vision_model.encoder.layers.5.layer_norm1.weight', 'vision_model.encoder.layers.5.layer_norm1.bias', 'vision_model.encoder.layers.5.self_attn.q_proj.weight', 'vision_model.encoder.layers.5.self_attn.k_proj.weight', 'vision_model.encoder.layers.5.self_attn.v_proj.weight', 'vision_model.encoder.layers.5.self_attn.out_proj.weight', 'vision_model.encoder.layers.5.layer_norm2.weight', 'vision_model.encoder.layers.5.layer_norm2.bias', 'vision_model.encoder.layers.5.mlp.fc1.weight', 'vision_model.encoder.layers.5.mlp.fc2.weight', 'vision_model.encoder.layers.6.layer_norm1.weight', 'vision_model.encoder.layers.6.layer_norm1.bias', 'vision_model.encoder.layers.6.self_attn.q_proj.weight', 'vision_model.encoder.layers.6.self_attn.k_proj.weight', 'vision_model.encoder.layers.6.self_attn.v_proj.weight', 'vision_model.encoder.layers.6.self_attn.out_proj.weight', 'vision_model.encoder.layers.6.layer_norm2.weight', 'vision_model.encoder.layers.6.layer_norm2.bias', 'vision_model.encoder.layers.6.mlp.fc1.weight', 'vision_model.encoder.layers.6.mlp.fc2.weight', 'vision_model.encoder.layers.7.layer_norm1.weight', 'vision_model.encoder.layers.7.layer_norm1.bias', 'vision_model.encoder.layers.7.self_attn.q_proj.weight', 'vision_model.encoder.layers.7.self_attn.k_proj.weight', 'vision_model.encoder.layers.7.self_attn.v_proj.weight', 'vision_model.encoder.layers.7.self_attn.out_proj.weight', 'vision_model.encoder.layers.7.layer_norm2.weight', 'vision_model.encoder.layers.7.layer_norm2.bias', 'vision_model.encoder.layers.7.mlp.fc1.weight', 'vision_model.encoder.layers.7.mlp.fc2.weight', 'vision_model.encoder.layers.8.layer_norm1.weight', 'vision_model.encoder.layers.8.layer_norm1.bias', 'vision_model.encoder.layers.8.self_attn.q_proj.weight', 'vision_model.encoder.layers.8.self_attn.k_proj.weight', 'vision_model.encoder.layers.8.self_attn.v_proj.weight', 'vision_model.encoder.layers.8.self_attn.out_proj.weight', 'vision_model.encoder.layers.8.layer_norm2.weight', 'vision_model.encoder.layers.8.layer_norm2.bias', 'vision_model.encoder.layers.8.mlp.fc1.weight', 'vision_model.encoder.layers.8.mlp.fc2.weight', 'vision_model.encoder.layers.9.layer_norm1.weight', 'vision_model.encoder.layers.9.layer_norm1.bias', 'vision_model.encoder.layers.9.self_attn.q_proj.weight', 'vision_model.encoder.layers.9.self_attn.k_proj.weight', 'vision_model.encoder.layers.9.self_attn.v_proj.weight', 'vision_model.encoder.layers.9.self_attn.out_proj.weight', 'vision_model.encoder.layers.9.layer_norm2.weight', 'vision_model.encoder.layers.9.layer_norm2.bias', 'vision_model.encoder.layers.9.mlp.fc1.weight', 'vision_model.encoder.layers.9.mlp.fc2.weight', 'vision_model.encoder.layers.10.layer_norm1.weight', 'vision_model.encoder.layers.10.layer_norm1.bias', 'vision_model.encoder.layers.10.self_attn.q_proj.weight', 'vision_model.encoder.layers.10.self_attn.k_proj.weight', 'vision_model.encoder.layers.10.self_attn.v_proj.weight', 'vision_model.encoder.layers.10.self_attn.out_proj.weight', 'vision_model.encoder.layers.10.layer_norm2.weight', 'vision_model.encoder.layers.10.layer_norm2.bias', 'vision_model.encoder.layers.10.mlp.fc1.weight', 'vision_model.encoder.layers.10.mlp.fc2.weight', 'vision_model.encoder.layers.11.layer_norm1.weight', 'vision_model.encoder.layers.11.layer_norm1.bias', 'vision_model.encoder.layers.11.self_attn.q_proj.weight', 'vision_model.encoder.layers.11.self_attn.k_proj.weight', 'vision_model.encoder.layers.11.self_attn.v_proj.weight', 'vision_model.encoder.layers.11.self_attn.out_proj.weight', 'vision_model.encoder.layers.11.layer_norm2.weight', 'vision_model.encoder.layers.11.layer_norm2.bias', 'vision_model.encoder.layers.11.mlp.fc1.weight', 'vision_model.encoder.layers.11.mlp.fc2.weight', 'vision_model.encoder.layers.12.layer_norm1.weight', 'vision_model.encoder.layers.12.layer_norm1.bias', 'vision_model.encoder.layers.12.self_attn.q_proj.weight', 'vision_model.encoder.layers.12.self_attn.k_proj.weight', 'vision_model.encoder.layers.12.self_attn.v_proj.weight', 'vision_model.encoder.layers.12.self_attn.out_proj.weight', 'vision_model.encoder.layers.12.layer_norm2.weight', 'vision_model.encoder.layers.12.layer_norm2.bias', 'vision_model.encoder.layers.12.mlp.fc1.weight', 'vision_model.encoder.layers.12.mlp.fc2.weight', 'vision_model.encoder.layers.13.layer_norm1.weight', 'vision_model.encoder.layers.13.layer_norm1.bias', 'vision_model.encoder.layers.13.self_attn.q_proj.weight', 'vision_model.encoder.layers.13.self_attn.k_proj.weight', 'vision_model.encoder.layers.13.self_attn.v_proj.weight', 'vision_model.encoder.layers.13.self_attn.out_proj.weight', 'vision_model.encoder.layers.13.layer_norm2.weight', 'vision_model.encoder.layers.13.layer_norm2.bias', 'vision_model.encoder.layers.13.mlp.fc1.weight', 'vision_model.encoder.layers.13.mlp.fc2.weight', 'vision_model.encoder.layers.14.layer_norm1.weight', 'vision_model.encoder.layers.14.layer_norm1.bias', 'vision_model.encoder.layers.14.self_attn.q_proj.weight', 'vision_model.encoder.layers.14.self_attn.k_proj.weight', 'vision_model.encoder.layers.14.self_attn.v_proj.weight', 'vision_model.encoder.layers.14.self_attn.out_proj.weight', 'vision_model.encoder.layers.14.layer_norm2.weight', 'vision_model.encoder.layers.14.layer_norm2.bias', 'vision_model.encoder.layers.14.mlp.fc1.weight', 'vision_model.encoder.layers.14.mlp.fc2.weight', 'vision_model.encoder.layers.15.layer_norm1.weight', 'vision_model.encoder.layers.15.layer_norm1.bias', 'vision_model.encoder.layers.15.self_attn.q_proj.weight', 'vision_model.encoder.layers.15.self_attn.k_proj.weight', 'vision_model.encoder.layers.15.self_attn.v_proj.weight', 'vision_model.encoder.layers.15.self_attn.out_proj.weight', 'vision_model.encoder.layers.15.layer_norm2.weight', 'vision_model.encoder.layers.15.layer_norm2.bias', 'vision_model.encoder.layers.15.mlp.fc1.weight', 'vision_model.encoder.layers.15.mlp.fc2.weight', 'vision_model.encoder.layers.16.layer_norm1.weight', 'vision_model.encoder.layers.16.layer_norm1.bias', 'vision_model.encoder.layers.16.self_attn.q_proj.weight', 'vision_model.encoder.layers.16.self_attn.k_proj.weight', 'vision_model.encoder.layers.16.self_attn.v_proj.weight', 'vision_model.encoder.layers.16.self_attn.out_proj.weight', 'vision_model.encoder.layers.16.layer_norm2.weight', 'vision_model.encoder.layers.16.layer_norm2.bias', 'vision_model.encoder.layers.16.mlp.fc1.weight', 'vision_model.encoder.layers.16.mlp.fc2.weight', 'vision_model.encoder.layers.17.layer_norm1.weight', 'vision_model.encoder.layers.17.layer_norm1.bias', 'vision_model.encoder.layers.17.self_attn.q_proj.weight', 'vision_model.encoder.layers.17.self_attn.k_proj.weight', 'vision_model.encoder.layers.17.self_attn.v_proj.weight', 'vision_model.encoder.layers.17.self_attn.out_proj.weight', 'vision_model.encoder.layers.17.layer_norm2.weight', 'vision_model.encoder.layers.17.layer_norm2.bias', 'vision_model.encoder.layers.17.mlp.fc1.weight', 'vision_model.encoder.layers.17.mlp.fc2.weight', 'vision_model.encoder.layers.18.layer_norm1.weight', 'vision_model.encoder.layers.18.layer_norm1.bias', 'vision_model.encoder.layers.18.self_attn.q_proj.weight', 'vision_model.encoder.layers.18.self_attn.k_proj.weight', 'vision_model.encoder.layers.18.self_attn.v_proj.weight', 'vision_model.encoder.layers.18.self_attn.out_proj.weight', 'vision_model.encoder.layers.18.layer_norm2.weight', 'vision_model.encoder.layers.18.layer_norm2.bias', 'vision_model.encoder.layers.18.mlp.fc1.weight', 'vision_model.encoder.layers.18.mlp.fc2.weight', 'vision_model.encoder.layers.19.layer_norm1.weight', 'vision_model.encoder.layers.19.layer_norm1.bias', 'vision_model.encoder.layers.19.self_attn.q_proj.weight', 'vision_model.encoder.layers.19.self_attn.k_proj.weight', 'vision_model.encoder.layers.19.self_attn.v_proj.weight', 'vision_model.encoder.layers.19.self_attn.out_proj.weight', 'vision_model.encoder.layers.19.layer_norm2.weight', 'vision_model.encoder.layers.19.layer_norm2.bias', 'vision_model.encoder.layers.19.mlp.fc1.weight', 'vision_model.encoder.layers.19.mlp.fc2.weight', 'vision_model.encoder.layers.20.layer_norm1.weight', 'vision_model.encoder.layers.20.layer_norm1.bias', 'vision_model.encoder.layers.20.self_attn.q_proj.weight', 'vision_model.encoder.layers.20.self_attn.k_proj.weight', 'vision_model.encoder.layers.20.self_attn.v_proj.weight', 'vision_model.encoder.layers.20.self_attn.out_proj.weight', 'vision_model.encoder.layers.20.layer_norm2.weight', 'vision_model.encoder.layers.20.layer_norm2.bias', 'vision_model.encoder.layers.20.mlp.fc1.weight', 'vision_model.encoder.layers.20.mlp.fc2.weight', 'vision_model.encoder.layers.21.layer_norm1.weight', 'vision_model.encoder.layers.21.layer_norm1.bias', 'vision_model.encoder.layers.21.self_attn.q_proj.weight', 'vision_model.encoder.layers.21.self_attn.k_proj.weight', 'vision_model.encoder.layers.21.self_attn.v_proj.weight', 'vision_model.encoder.layers.21.self_attn.out_proj.weight', 'vision_model.encoder.layers.21.layer_norm2.weight', 'vision_model.encoder.layers.21.layer_norm2.bias', 'vision_model.encoder.layers.21.mlp.fc1.weight', 'vision_model.encoder.layers.21.mlp.fc2.weight', 'vision_model.encoder.layers.22.layer_norm1.weight', 'vision_model.encoder.layers.22.layer_norm1.bias', 'vision_model.encoder.layers.22.self_attn.q_proj.weight', 'vision_model.encoder.layers.22.self_attn.k_proj.weight', 'vision_model.encoder.layers.22.self_attn.v_proj.weight', 'vision_model.encoder.layers.22.self_attn.out_proj.weight', 'vision_model.encoder.layers.22.layer_norm2.weight', 'vision_model.encoder.layers.22.layer_norm2.bias', 'vision_model.encoder.layers.22.mlp.fc1.weight', 'vision_model.encoder.layers.22.mlp.fc2.weight', 'vision_model.encoder.layers.23.layer_norm1.weight', 'vision_model.encoder.layers.23.layer_norm1.bias', 'vision_model.encoder.layers.23.self_attn.q_proj.weight', 'vision_model.encoder.layers.23.self_attn.k_proj.weight', 'vision_model.encoder.layers.23.self_attn.v_proj.weight', 'vision_model.encoder.layers.23.self_attn.out_proj.weight', 'vision_model.encoder.layers.23.layer_norm2.weight', 'vision_model.encoder.layers.23.layer_norm2.bias', 'vision_model.encoder.layers.23.mlp.fc1.weight', 'vision_model.encoder.layers.23.mlp.fc2.weight', 'vision_model.encoder.layers.24.layer_norm1.weight', 'vision_model.encoder.layers.24.layer_norm1.bias', 'vision_model.encoder.layers.24.self_attn.q_proj.weight', 'vision_model.encoder.layers.24.self_attn.k_proj.weight', 'vision_model.encoder.layers.24.self_attn.v_proj.weight', 'vision_model.encoder.layers.24.self_attn.out_proj.weight', 'vision_model.encoder.layers.24.layer_norm2.weight', 'vision_model.encoder.layers.24.layer_norm2.bias', 'vision_model.encoder.layers.24.mlp.fc1.weight', 'vision_model.encoder.layers.24.mlp.fc2.weight', 'vision_model.encoder.layers.25.layer_norm1.weight', 'vision_model.encoder.layers.25.layer_norm1.bias', 'vision_model.encoder.layers.25.self_attn.q_proj.weight', 'vision_model.encoder.layers.25.self_attn.k_proj.weight', 'vision_model.encoder.layers.25.self_attn.v_proj.weight', 'vision_model.encoder.layers.25.self_attn.out_proj.weight', 'vision_model.encoder.layers.25.layer_norm2.weight', 'vision_model.encoder.layers.25.layer_norm2.bias', 'vision_model.encoder.layers.25.mlp.fc1.weight', 'vision_model.encoder.layers.25.mlp.fc2.weight', 'vision_model.encoder.layers.26.layer_norm1.weight', 'vision_model.encoder.layers.26.layer_norm1.bias', 'vision_model.encoder.layers.26.self_attn.q_proj.weight', 'vision_model.encoder.layers.26.self_attn.k_proj.weight', 'vision_model.encoder.layers.26.self_attn.v_proj.weight', 'vision_model.encoder.layers.26.self_attn.out_proj.weight', 'vision_model.encoder.layers.26.layer_norm2.weight', 'vision_model.encoder.layers.26.layer_norm2.bias', 'vision_model.encoder.layers.26.mlp.fc1.weight', 'vision_model.encoder.layers.26.mlp.fc2.weight', 'vision_model.post_layernorm.weight', 'vision_model.post_layernorm.bias']
[2026-07-01 19:29:35.483] CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
[2026-07-01 19:29:35.568] Requested to load LTXAVTEModel_
[2026-07-01 19:29:36.646] FETCH ComfyRegistry Data: 135/157
[2026-07-01 19:29:41.215] FETCH ComfyRegistry Data: 140/157
[2026-07-01 19:29:46.067] FETCH ComfyRegistry Data: 145/157
[2026-07-01 19:29:51.164] FETCH ComfyRegistry Data: 150/157
[2026-07-01 19:29:56.097] FETCH ComfyRegistry Data: 155/157
[2026-07-01 19:29:58.977] FETCH ComfyRegistry Data [DONE]
[2026-07-01 19:29:59.397] [ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
[2026-07-01 19:29:59.613] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[2026-07-01 19:30:00.221] [ComfyUI-Manager] All startup tasks have been completed.
[2026-07-01 19:30:15.596] loaded partially; 7087.88 MB usable, 0.00 MB loaded, 15964.13 MB offloaded, 7966.42 MB buffer reserved, lowvram patches: 0
[2026-07-01 19:30:15.630] Attempting to release mmap (290)
[2026-07-01 19:30:30.925] loaded partially; 7030.75 MB usable, 0.00 MB loaded, 15964.13 MB offloaded, 7966.42 MB buffer reserved, lowvram patches: 0
[2026-07-01 19:30:34.669] Requested to load VideoVAE
[2026-07-01 19:30:34.718] Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
[2026-07-01 19:30:36.242] Found quantization metadata version 1
[2026-07-01 19:30:36.242] Detected mixed precision quantization
[2026-07-01 19:30:36.247] Using mixed precision operations
[2026-07-01 19:30:36.247] Native ops: , emulated ops: mxfp8, float8_e5m2, nvfp4, float8_e4m3fn
[2026-07-01 19:30:36.382] model weight dtype torch.bfloat16, manual cast: torch.bfloat16
[2026-07-01 19:30:36.402] model_type FLUX
[2026-07-01 19:30:38.098] VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
[2026-07-01 19:30:38.102] no CLIP/text encoder weights in checkpoint, the text encoder model will not be loaded.
[2026-07-01 19:30:39.289] Sampling split indices: [3, 6]
[2026-07-01 19:30:39.367] Sigmas chunks: [tensor([1.0000, 0.9937, 0.9875, 0.9812]), tensor([0.9812, 0.9750, 0.9094, 0.7250]), tensor([0.7250, 0.4219, 0.0000])]
[2026-07-01 19:30:39.367] Sampling with sigmas tensor([1.0000, 0.9937, 0.9875, 0.9812])
[2026-07-01 19:30:39.479] Requested to load LTXAV
[2026-07-01 19:30:42.562] Model LTXAV prepared for dynamic VRAM loading. 24203MB Staged. 1248 patches attached. Force pre-loaded 2070 weights: 3308 KB.
[2026-07-01 19:32:40.072]
100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [01:06<00:00, 19.02s/it]
100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [01:06<00:00, 22.21s/it]
[2026-07-01 19:32:40.210] After 3 steps, the latent image was normalized by 1.000000 and 0.250000
[2026-07-01 19:32:40.210] Sampling with sigmas tensor([0.9812, 0.9750, 0.9094, 0.7250])
[2026-07-01 19:32:42.039] 0 models unloaded.
[2026-07-01 19:32:42.857] Model LTXAV prepared for dynamic VRAM loading. 24203MB Staged. 1248 patches attached. Force pre-loaded 2070 weights: 3308 KB.
[2026-07-01 19:34:41.142]
100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [01:18<00:00, 20.97s/it]
100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [01:18<00:00, 26.33s/it]
[2026-07-01 19:34:41.269] After 6 steps, the latent image was normalized by 1.000000 and 0.250000
[2026-07-01 19:34:41.269] Sampling with sigmas tensor([0.7250, 0.4219, 0.0000])
[2026-07-01 19:34:43.159] 0 models unloaded.
[2026-07-01 19:34:43.946] Model LTXAV prepared for dynamic VRAM loading. 24203MB Staged. 1248 patches attached. Force pre-loaded 2070 weights: 3308 KB.
[2026-07-01 19:37:02.444]
100%|████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:59<00:00, 29.61s/it]
100%|████████████████████████████████████████████████████████████████████████████████████| 2/2 [00:59<00:00, 29.61s/it]
[2026-07-01 19:37:09.923] After 8 steps, the latent image was normalized by 1.000000 and 1.000000
[2026-07-01 19:37:56.355] Model VideoVAE prepared for dynamic VRAM loading. 1384MB Staged. 0 patches attached.
[2026-07-01 19:38:08.647] Sampling split indices: [3, 6]
[2026-07-01 19:38:08.647] Sigmas chunks: [tensor([0.9094, 0.7250, 0.4219, 0.0000]), tensor([0.])]
[2026-07-01 19:38:08.652] Sampling with sigmas tensor([0.9094, 0.7250, 0.4219, 0.0000])
[2026-07-01 19:38:09.170] 0 models unloaded.
[2026-07-01 19:38:09.907] Model LTXAV prepared for dynamic VRAM loading. 24203MB Staged. 0 patches attached. Force pre-loaded 2070 weights: 3308 KB.
```

### Other

After last entry in log both System RAM and VRAM are full and zero activity on both CPU and GPU. Machine unresponsive and requires forced reboot on either Windows 11 or CachyOS.

Local Solution: Patched ComfyUI/comfy/quant_ops.py
import os
if os.environ.get("COMFY_KITCHEN_DISABLE_CUDA") == "1":
ck.registry.disable("cuda")
logging.warning("COMFY_KITCHEN_DISABLE_CUDA=1: disabled comfy-kitchen CUDA backend for diagnostics.")

Then I added an environment variable in the batch file:
set COMFY_KITCHEN_DISABLE_CUDA=1

Start with the ComfyUI/main.py launch configuration and the comfy-kitchen==0.2.10 CUDA backend, then reproduce the repeated LTX 2.3 queue using the command and settings in the report. Compare the logs around “Model LTXAV prepared for dynamic VRAM loading” with the backend-disabled run. Done means repeated generations no longer hard-freeze the host and failures are reported or recovered by ComfyUI/PyTorch.

Written by the indexing model from the issue text.