[Feature Request]: My suggestions and a problem report regarding Head Swap V1 Flux 2 Klein 4b_9b

[Eva-033]

Description

1. Insert acceleration nodes like tea cache or nunchaku (effective for Flux) to accelerate LoRa using Turbo;
2. Add nodes like cleanGpuUsed to clean up VRAM;
3. Connect nodes like Fast Muter (rgthree) to manage which nodes are muted or enabled.
4. Negative warning words related to Flux2 Klein should be ineffective (especially when CFG is set to 1). Generally, adding a ConditioningZeroOut node to skip them is faster and less prone to errors. My understanding comes from the official website and other developers; I don't fully understand the principles, so please forgive me if I'm wrong.

I encountered a problem during use: when running Head Swap V1 Flux 2 Klein 4b_9b in the latest version of Comfyui, it gets stuck at the final VAEDecode node. Switching to the desktop and refreshing skips it. This happens frequently with 9B+LoRa. However, this doesn't happen when others use your workflow mix (zit+face swap). How can this be resolved?

Additional information

No response

[Eva-033]

I need to add an additional question. Every time I close the workflow, I am prompted to save, even if I haven't made any changes. How can this issue be resolved? Through step-by-step testing using the binary method, I discovered that the problem occurs at the ResizeImageMaskNode node. It may conflict with my comfyui or plugins. I tried upgrading this node to the latest version, but it didn't work. I don't know how to solve this problem. Can you provide help?
With the help of AI, I replaced the original node with the MaskResize node from ComfyUI Essentials. Although this solved the problem, it is not ideal, because for others who encounter the same issue, it may require installing the Essentials node package instead of using the native node.