I definitely think this is a cool idea, but there are potentially a few issues I can think of:
I think this is a cool idea, too. For a response:
Assuming you were ever going to have small icons for games when navigating through a user's library, you could fall back on that as a default. Maybe you could just apply that texture to a plane for games that don't have such a model defined.
You probably could get away with it alright if you restrict the number of models visible at any given time and the number of polygons visible in the 3D figure.
> Not all games are 3d, and it might be difficult to have some creators make 3d assets for their game
There are few fallback options: the first simplest one is to use a character in a 2d transparent (with no square border using alpha channel) which may be simple (one simple sprite) or more elaborated (different images for sprite being watch from front, side or back). Voxel art is another option: there are plenty of free editor and stuff like that, and should be accessible for anyone (think small characters like the one in rpgmaker)
> From an accessibility standpoint it's harder to navigate a 3d space
3D space offer a wide range of options that can be simply activate/deactivated with one simple click. From 3D you can always return to seamlessly "return to 2D" using the same data (think Fez game) while in 2D the best you can do to achieve 3D is the usual "cover flip". Basically 3D can generate a real 2D experience (ie: lock the camera to 2 axis and a simple zoom to have far/close view) and then enable customers to a real First Person experience. In 2D all you can have is the " flip cover", infinite scroll on XY (like google maps, minus the real 3D option) and basic scrolling. In 2D transition between these modes require additional work for every feature... while in 3D is just about lock certain camera's axis
> Performance issues, if you have a lot of games visible at once and someone is on a slow laptop or phone then it might make browsing difficult
I can theorize few options here: the first and most simple one is just about use the classical "fog" trick, in which the viewer can set up the distance based upon their own experience/hardware. General optimization can be the use of Level Of Detail technique, restrict the model to only a limited amount of vertex and UV size. Imposter billboard can also be auto generated: there are tools and guides for it.
Mobile app would be still possible (with even more aimed optimizations) but it would be a bit of a overkill: not even Steam's Big Picture mode is available for mobile, and that's even a 2D app which emulate the "simplicity of mobile ".
Anyway, a 3D store integrated in an Android TV box? That would be a cool idea tho.
