Fb ML software engineer vs. research scientist

I noticed some teams at fb have both ML software engineers and research scientists working together. What are the major differences btw. these two roles? For example in terms of type of work (RS focuses more on state-of-the-art papers, only building proof of concept prototypes without worrying about production, whereas SWE works more on actual product not fancy academic publications) and performance evaluation metrics (RS is evaluated by publications whereas SWE is evaluated on user metric gains). Are the lines more blurred than I think?