Hybrid simulation can provide significant advantages for large-scale experimental investigations of the seismic response of structures through collapse, particularly when considering cost and safety of conventional shake table tests. Hybrid simulation, however, has its own challenges and special attention must be paid to mitigate potential numerical and experimental errors that can propagate throughout the simulation. Several case studies are presented here to gain insight into the factors influencing the accuracy and stability of hybrid simulation from the linear-elastic response range through collapse. The hybrid simulations were conducted on a four-story two-bay moment frame with various substructuring configurations. Importantly, the structural system examined here was previously tested on a shake table with the same loading sequence, allowing for direct evaluation of the hybrid simulation results. The sources of error examined include: (1) computational stability in numerical substructure; (2) setup and installation of the physical specimen representing the experimental substructure; and (3) the accuracy of the selected substructuring technique that handles the boundary conditions and continuous exchange of data between the subassemblies. Recommendations are made regarding the effective mitigation of the various sources of errors. It is shown that by controlling errors, hybrid simulation can provide reliable results for collapse simulation by comparison to shake table testing.