> Science is a system

Indeed; journals are repos, authors are committers, reviews are pull requests. But the unit tests are insufficient, and literature references don't quite make it as a form of continuous integration.