Multidisciplinary Design Optimization frameworks and process integration software in general are an active R&D area. Commercial vendors keep developing platforms, and the open-source community is continuing to develop well-established software such as OpenMDAO, MDAx and GEMSEO, while also exploring alternatives such as the recent Amigo, an HPC and GPU-oriented framework. Although they follow different principles in terms of the MDO process configuration, execution and monitoring, their core features are based on optimisation, the chain rule and the adjoints or the MAUD method for handling derivatives, design of experiments and so on. This provides a solid foundation for implementing more advanced strategies that include and combine multi-fidelity, optimisation preconditioning, machine learning, multi-objective and robust MDO. In this presentation, we will first introduce the recent advances in the GEMSEO framework with regard to these capabilities. We will then focus on the open scientific challenges arising from these capabilities and their combination : • accelerating MDO processes by reusing information from previous executions, particularly preconditioning of optimisation solvers, • the combination of multi-objective, bi-level and multi-fidelity MDO, • the theoretical aspects of MDO formulations and their link with the convergence properties of algorithms, • the creation of MDO benchmark problems with the best balance between representing real applications and the CPU cost required to solve them.