In software language engineering, composition and modular design are essential milestones to advance reuse in this domain. Although language engineering improves quality and development efficiency by incorporating reusable components and tooling, some disadvantages and limitations can inhibit sophisticated language development. Current research usually focuses on the conceptual advantages but neglects the often technical drawbacks resulting from the necessity for complex interactions for seamless integration. In this paper, we report on our experiences applying language composition with the language workbench MontiCore and elaborate on the conceptual and technical drawbacks of the intended compositional tooling. Using concise application examples, we demonstrate the challenges of a compositional parser, the complexity of the generated infrastructure, and how they scrape the limits of the underlying target programming languages and their compilers. This critical examination of the implications of modular language construction demonstrates the pitfalls of language composition in the large and should reveal potential for future research.