"Someone I know" is Puppy Games (http://www.puppygames.net) and they get a 65% success rate for OpenGL on Windows. 35% *is* a massive failure rate. Totally unacceptable.
What is their solution then? I was not able to find it on thier site, nor was I able to confirm the failure rate. Using third party libs before they've been proven is a dangerous game.
Were that true, new technology would never have a chance to be proven and people would still be using DirectX 3 and software renderers. Nothing ventured, nothing gained. Mark has the Blitz3D engine, which presumably a DirectX module would be based on, which *is* a proven stable and compatible engine.
You forgot "that as far as we know might blow up whhen applied to the BlitzMAX framework". Unless you're advocating sticking to Blitz3D for now, which I wholeheartedly agree with, if you need DirectX. Another proven engine is the Popcap game framework. That is a real possibility.
Not for 3D. In either case the popcap framework gets much worse performance on my TabletPC than BlitzMAX, so it's not in itself a magic bullet solution for everyone. Where does religion come into it? This *is* about business.
What's the problem then? I hate discussions like this, but it seems the impact is a big one.
I like them. If there really is a problem, then it's important to uncover it early on.