Games Domain has put up another preview (http://gamesdomain.yahoo.com/preview/40442) of Dragonshard along with a new video showing some new game footage while Liquid president Ed Del Castillo gives his description of the game. One thing to note is that Games Domain operates with 03/15/2005 as a release date, though that may be speculations. Here's a clip from the preview:

Basic overworld units include rare, high-powered heroes, captains, and grunt soldiers. In a big departure from convention, Dragonshard does away with the tech tree. Every unit, piece of "technology," and building is available right from the start. Instead of requiring players to go through specific steps to obtain units and buildings, the game rewards strategic thinking in how each is deployed. For example, it's possible to create heavily armored units by placing a blacksmith next to the barracks that produce them. Liquid promises thousands of possible results from experimenting with how building placement and units are combined.

I don't know what's best: a really big manual or exploring the thousands of possibilites myself?

Considering that their primary target audience are D&D players who usually only play CRPGs, having "thousands of possibilities" for anything will only mean people will think the game is way too complicated for them to bother with. Sounds to me like they made things unnecessarily complicated just so they can advertise the game as "better" or more advanced than other RTS. Unless, of course, other RTS players pick the game up, but they might not want to because they don't know anything about D&D rules. But then again, they'll probably be so dumbed down in the game you won't need to know anything to be able to play.

I really wonder how this thing will turn out.

My guess is that the thousands of possibilities for the units come down to only a hundred (if that many) combinations between buildings. Unitproducing buildings have a tendency to build more than one unit in strategy games, and hence, one combination of a production building with another building could "upgrade" all of the produced units, hence making for alot of units resulting from a minimum of building combos.