Whilst I must admit that I have never done a proper crowd scene, I think the answer is to use some kind of layering. Rendering 1,2,3, or more characters at a time, then using that render as a 2D element in the next render, to which two (for example) new characters are added. There are various ways to integrate the 2D images. Pasting them into square primitives  in the scene, or pasting them into the Poser background, or assembling them in a paint program as layers, or a combination of the above.

Which ever method is used, careful consideration needs to be given to what will be included in each layer, and what things in one layer will occlude elements in other layers. Shadows can be a problem in these methods.