The assessment of the seismic performance of unreinforced masonry cross-vaults is still a challenge in numerical analysis, due to complex curved geometries and bond patterns, and uncertainties related to the selection of adequate modeling strategies, including but not limited to that of material properties, damping scheme, and unit/joint idealization. This paper presents the results of a collaborative effort to validate, against the shake table test of both unstrengthened and strengthened masonry cross-vault specimens as part of the SERA Project Blind Prediction and Post-diction Competition, various discontinuum-based numerical approaches. First, the geometry of the cross-vault is created using a Python-based computational framework to accurately represent the brick arrangement and the shape of the vault. Then, the geometry is converted into an assemblage of deformable blocks and analyzed using the Distinct Element Method (DEM). An elasto-softening contact model based on fracture energy is implemented in the masonry joints to simulate crushing, tensile, and shear failures. The performance of the proposed strategy, conceived for the unstrengthened configuration of the tested vault specimen and then adapted to include the presence of cementitious repairs, shows satisfactory agreement with both qualitative and quantitative experimental responses, also revealing critical insights and lessons learned through the blind/post-prediction exercise.