The reliability of different parameters in the surface hopping method is assessed for a vibronic coupling model of a challenging transition metal complex, where a large number of electronic states of different multiplicities are met within a small energy range. In particular, the effect of two decoherence correction schemes and of various strategies for momentum rescaling and treating frustrating hops during the dynamics is investigated and compared against an accurate quantum dynamics simulation. The results show that surface hopping is generally able to reproduce the reference but also that small differences in the protocol used can strongly affect the results. We find a clear preference for momentum rescaling along only one degree of freedom, using either the nonadiabatic coupling or the gradient difference vector, and trace this effect back to an enhanced number of frustrated hops. Furthermore, reflection of the momentum after frustrated hops is shown to work better than to ignore the process completely. The study also highlights the importance of the decoherence correction, but neither of the two methods employed, energy based decoherence or augmented fewest switches surface hopping, performs completely satisfactory and we trace this effect back to a lack of size-consistency. Finally, the effect of different methods for analyzing the populations is highlighted. More generally, the study emphasizes the importance of the often neglected parameters in surface hopping and shows that there is still a need for simple, robust, and generally applicable correction schemes.