The principal incentive of science is publish or perish. Novelty trumps replication of the previous results (see replication crisis). Use of the citation metric

The principal incentive of science is publish or perish. Novelty trumps replication of the previous results (see replication crisis). Use of the citation metrics to evaluate performance in science is similar to relying on likes in the social media to measure importance. The number of citations matter more than the effort (cost) of producing the result. This environment favors those who can deliver something that's good enough, and then move on (and publish often). Software engineering is a continuous effort. Maintenance often requires more time and resources than the original development. It's more like curating a library or a database. Their product is perceived to stay the same. Incremental improvements do not advance the career of the maintainer at all. Researchers who sink their time into software development mostly hurt their academic careers. Even if they publish a paper about ResearchTool 1.0.0, they won't be able to publish another one about ResearchTool 1.0.1 and another one about ResearchTool 1.0.2. Meanwhile, their peers will probably publish "A", "B", and "C". Science needs a career path which does not depend on the number of publications.