Concern for lack of information and backwards compatibility across repackaged uploads.
I think uploading "repackaged" versions of models for a better user experience down the line for ComfyUI is a good idea, though I'd find it very helpful if the ReadMe stated what, if anything, is different from the original model.
This model in particular seems to have the text _model/text _projection part removed since I assume it is never used in ComfyUI (or generally anywhere probably), which results in smaller filesize and resource usage, which is nice, however should a custom node ever require the text encoder part of the model the user would likely be very confused why the custom nodes don't work and the developer of them may not be aware of this repack either, and would have to inspect the model first to even realize that the text part is missing.
Of course, I am fully aware that it's very unlikely for the text part to ever be used, especially for this clip vit h model. But it would still be very helpful if the ReadMes on repackaged model uploads just contained this simple note of what is different (text part removed since seems unnecessary), and why (reduced file size, lower resource usage).
Often these repackaged models seem to be not backwards compatible as is with other backends or even diffusers
For Z-Image-Turbo repackaged the DiT itself isn't backwards compatible with diffusers, but at least there's a script added to show what was changed. But there, the best case I think would be for comfyui to just use and support the original keys and not any renamed ones