I have achieved a similar thing using a copy of QlikView Desktop and the API. Write some VBscript which cycles around some of your common listboxes. It can then be scheduled as a Supporting Task in Publisher following your Reload and Distribution task.
i'am during model perfection. But with this data volume each mutation (all almost clear stars, minimal use of set analysis) without cache response time is on the 30 s. level. With ready cache it is immediately.
Of course i know, that it is impossible to prepare cache with all potential users selection states. I think, cache will be very big, but (using pareto) probably 20% of cache will have 80% important users selections states. And this i why it is important to know, what users will be looking for in your big app.