Is it feasible to use half-edge mesh in games instead of vertex-face structure? What comes to my mind is that it could be useful in mesh simplification (dynamic level of detail). But I'm worried it could take too much memory or even have a negative impact on game's rendering performance.
I just read about the half-edge mesh structure; this page was helpful. Thanks for showing it to me!
If your main game mechanic involves lots of dynamic level of detail calculations on meshes, then it would be a great choice. Also if you were developing a mesh editor, you might consider using it if the user would often perform operations which the structure is better at. Otherwise, it doesn't seem like an advantage for game purposes. Personally, I would rather precalculate level of detail meshes and then use the conventional mesh format.
Furthermore, the restriction that it can only represent manifold surfaces sounds like a deal-breaker to me. I imagine very few game models are actually capable of being represented this way.
I think with any alternate data structure, you should consider the cons along with the pros. This data structure looks neat, it certainly has its pros, but it doesn't seem a better fit for games than the current widely used format of vertices and faces (whatever its name may be).