Thank you for referencing Tom Fawcett’s article, which can be found here: https://www.math.ucdavis.edu/~saito/data/roc/fawcett-roc.pdf.

He made the same point that ROC cannot be an effective tool when there is class skew. He compares a balanced class scenario with an imbalanced class scenario in Figure (5) of the article. The ROC curves of the two scenarios are identical, but the precision-recall curves are dramatically different. He said “…observe that the ROC graphs in Fig. 5a and c are identical, while the precision-recall graphs in Fig. 5b and d differ substantially.”

So check it out!