I was told in #10748 (comment) that there shouldn't be too many image comparison tests, and I assume (though I didn't ask) that this is primarily because the repository size would grow too large, in particular as these images have to be replaced whenever significant changes to the code are made. Are there other important reasons that I'm missing?

Anyway, I wanted to ask: has there ever been discussion about tracking images for image comparison tests via git-lfs? It would only store a hash of each image in the repository, while the actual file contents are hosted separately, so that e.g. cloning the repo can be sped up by only downloading a few recent image versions (of course, it makes absolutely no difference if you really need to have the entire history of all images locally).

Thoughts? Couldn't find any existing issues mentioning LFS, sorry if this has already been discussed elsewhere.