Model-driven software engineering has gained momentum in academia as well as in industry for improving the development of evolving software by providing appropriate abstraction mechanisms in terms of software models and transformations thereof. With the rise of cyber-physical systems in general, and cyber-physical production systems in particular, the interplay between several engineering disciplines, such as software engineering, mechanical engineering and electrical engineering, becomes a must. Thus, a shift from pure software models to cross-disciplinary models has to take place to develop the full potential of model-driven engineering for the whole production domain. Cross-disciplinary models are also essential to raise the level of flexibility of production systems in order to better react to changing requirements, since systems are no longer designed to be, but they have to be designed to evolve. In this talk, we will have a look at current practice of good, bad, and ugly cross-disciplinary modeling. We will point to ongoing work of (hopefully) improving this situation by applying and further developing model-driven techniques such as consistency management and co-evolution support for the production domain.