Not really weighing in one way or another, but I do not see scatter plots in the paper other than one PCA plot (which was not LDA... not sure why they chose PCA

Not really weighing in one way or another, but I do not see scatter plots in the paper other than one PCA plot (which was not LDA... not sure why they chose PCA, so no strong reason to expect point separation there, though I do not know the diversity metric they are using so feel free to ignore my commentary). Violin plots always look like they overlap. I think seaborn's split=True should be pushed a bit harder.