Going into the metric they're using, it feels like 'it's complicated'. They're saying that the disruptiveness of work is decreasing, and they measure disruptiveness in a certain way: the degree to which future papers cite a given focal work but also stop citing the predecessors of that focal work.

That's not necessarily the same as saying that the pace of discovery is in decline. It might be saying that the rate of *transformative* discovery is in decline. Or it might just be saying that there are now so many papers that any individual paper just isn't getting picked up consistently enough to transform a field on its own, and the transformative objects now might be sets of connected work rather than a single revolutionary discovery - which would be my bet (especially given evolving lab practices and the emergence of the 'minimal publishable unit' style of work).