Link to the posts showing my CLF preview tier edge case issues, as brought up in the meeting, for those who are interested.
But just to emphasise that these results indicate that it is hard to empirically evaluate the quality of a baked 3D LUT vs multi-step ground truth. It does not suggest that it is hard to bake a complex CLF into a simple 3D LUT (indeed it’s pretty trivial) or that people should not do so for preview purposes. As @ptr mentioned, that works fine for the purposes for which it is intended.