Somic Xtreme is a sad tail of many things going wrong and many people believing in different goals. There are so many different parts in the story, making a big drama. I feel sad for reading it, though it was obvious that the initial idea of two seperated game modes, combined to one game, was a wrong step in the beginning.
As sad as the whole tale is, it's even more sad that I've never a game using this kind of fish-eye lense. Personally, I'd love to play a game in this kind of perspective. It's completely refreshing and offers the possibilities to show and hide map contents immediately based on the level design. I'm amazed by this spherical mode. I can only guess how it works in detail by looking at leaked game or editor screenshots. Seems to be some kind of threedimensional grid converted to 3D using the fisheye calculation and then rendered normally. And looking at this video I watched several times, I'd say at position 2:15 you can get a good shot at how the camera renders the tiles. It's not a pretty shot, but shows off how effective it can handle even tiles with relatively big distances from the camera. I'd give my left arm for talking or with this developer. And this video makes me quite about the seperation between conventional, hardware-accelerated 3D and software-based fisheye coordinate transformation. You can pretty often see how the distance clipping is done on a per-tile distance level independent from the fish-eye effect. So he must have checked visibility only by looking at it's tiles. And in the first video (at 2:15 again) you can also see edges on the tiled walls, indicating that they are rendered as normal polygons, not completely round fisheye-translated objects. I haven't yet seen a full rotation of the sphere with repeating map data, so I'm 100% sure it worked this way. Even repeating maps are possible when including such stuff in the renderer. This'd be only a minor tweak in the system as far as I can guess.
Quite clever. You don't need to make any complicated 3D-space distance calculation, just basic onedimensional plus/minus formulas. And knowing how much the hardward can handle/how many tiles could be one the screen, this makes a formidable technical archievement.
And yes, of course could you say it must be a conversation to normal 3D. How else could a simple polygon-based graphics card handle all this stuff without more advanced models like shaders. A shame the two responsible developers could never make a real game out of it. You should totally read the Wikipedia article linked in the first sentence. It gives some very intense view into how bad development diversion can go.
Edit Nr. 1: There's a FAQ by one of the developers. It could give me some insight and what they all talk about the editor, but not the engine. Well, I'll take a look and make another post if there's something worthwhile I can get out of it.
Edit Nr.2: The FAQ doesn't tell me anything about the engine background but about the other, non-programming details. But it seems that the editor was kinda the complete game editor. Including sound effects, enemy movement, level data of course, and so on. So yeah, he did all the hard work. And again, it's a shame I can't learn more about it. I don't want any of the resulting works, just more information. The features mentioned are more than just feature. It's some kind of totally new game concept you can develop with it - no doubt that it would've taken a while or two more to produce something final with it.
Why do corporations and alike often bounce so awesome concepts? In this case, Sega had some serious business problems, of course. Excluding this fact rather leads to confusion and not even excuse. I'm kinda glad that Nintendo often tries to incorporate new ideas and concepts (though I strongly dislike some of them, but that doesn't count here). Another reason to support them and buy their video games, atleast from my point of view.