This is a justified question. As you can see in the forecast page, multiple bones per vertex are planned, however it's not good to rely such a decision on future features.

I don't think that one bone per vertex gives such a bad result in a realtime game as your artist suggests. You could maybe do a test with one model.

The only other alternative at the moment is vertex animation. It is even faster than bones animation, but the drawback is memory requirement. You can calculate whether vertex animation is practical for your model: If you use a 10000 vertex model, every frame consumes 120 KB virtual memory. For 100 frames, this is 12 MB. If you don't use too many different models of this size, it is still practical. The virtual memory requirement of a huge level is in the 150..200 MB range.

I don't now much about Cal3D, but as far as I understand this would force the user to only use this proprietary format for importing. This does not seem like a good solution. The import format what we intend to implement next is FBX rather than Cal3D.