It is not a mandatory step, but in this case it could help in the alignment process. In general, images with the same body/optics must have the same or very similar calibration parameters. These parameters may vary due to small movements of optics caused by focusing, mechanical movement or temperature changes. These differences are negligible for small resolution cameras, while for high resolution cameras, e.g. 60 Mpx and more, these differences are more visible. On the other hand, by grouping calibration parameters together RealityCapture will need to calculate less parameters which results in a faster computation.
But not doing that shouldn’t cause any crashes.