In my view, a !D model for natural rivers needs to treat vegetation, both rigid and specially flexible vegetation.
Furthermore, the shear stress distribution on the bed should be carefully modeled: I mean, above all, to model the friction between the river main channel and its floodplains.
Bedload transport by fractions is another important function to be included.

Proper sediment transport and morphological processes, equivalent to Sobek-RE, will already be good enough. Besides, some (user-friendly) functionalities in 1D2D model, e.g. dam-break features given that the 1D2D model is widely used for dam-break computations and analysis (check HEC-RAS, for example).