Poster at IMDIS 2024, session tech, poster number 84

We employ standard continuos integration tools in order to monitor the robustness of important data products to changes in the software environment. We argue that trust in data products relies on trusting both the scientific methods and their software implementations. In both cases correctness cannot be proven, but trust can be gained by monitoring the existing processes of critical evaluation to check if correctness is corroborated. We extend on existing transparency standards that provide reproducibility in the sense of facilitating exact recomputation of data products, to a richer and more classical sense of reproducibility that verifies that data products are insensitive to the details of the software environment that they have implicitly been assumed to be insensitive of.