In this talk, we will discuss the ideal of reproducible science through the lens of interdisciplinary research within the natural sciences. In particular, we will consider the notion of "reproducible" within the context of experimental research, classical computational science, and computational approximations involving some data-driven or stochastic component. While the talk will try to stimulate a discussion of these topics in general, it will be rooted in the authors experiences within the "PoroTwin"-project, which aims to create an autonomous FluidFlower-laboratory by leveraging state-of-the-art digital twin concepts.