Recent mesoscale experimental observations of incipient spall failure, e.g. (Brown et al., 2015), have demonstrated a strong relationship between grain boundary misorientation and the likelihood of failure initiation along said grain boundary. This relationship has been attributed to an inherent weakness and disorder of grain boundaries of particular misorientation. Here we discuss the role played by mechanics, e.g. elastic and slip anisotropy effects, on this phenomenon. We make use of a recently developed framework for modeling dislocation-based crystal plasticity and ductile failure of single crystals under dynamic loading (Nguyen et al., 2017). Polycrystalline samples are studied at the mesoscale level through the explicit resolution of each grain, i.e. resolving each individual grain size, shape, and orientation, in a representative volume element. In our simulations, failure naturally localizes along grain boundaries with no necessity for ad hoc rules governing damage nucleation. We carry out a few hundred mesoscale calculations, systematically varying the misorientation angles of the grain boundaries in the computational microstructure. Despite the fact that we neglect the possibility of variations in inherent grain boundary weakness, our simulations agree favorably with the experimental observations of Brown et al., 2015. Furthermore, we find that misorientation angle is an insufficient description of the grain boundary for such correlations. In particular, we find that it is necessary to consider all five degrees of freedom to characterize each grain boundary, with tilt angles being of greatest influence. Lastly, we discuss some implications of our analysis on the anomalous grain size dependence, c.f. (Wilkerson and Ramesh, 2016), of spall strength.