Or why you can't have 600 seperate entities
That's strange. My old geforce 2 runs a blitz3D app at well over 40 FPS (which is good for a geforce 2) while drawing 6000 individual entities.
Once you start adding gameplay, physics, collision, AI. Start trying to optimize culling and LOD and use fairly advanced materials you get a lot of limitations in how you can organise your scenes and draw them.
Gameplay? Physics? That's up to the programmer to do. If you can't make a fun game, please don't blame it on Blitz3D.
Colling and LOD I have done many times with extremely good results, I've even had about 1000 * 1000 poly models (all on the screen at the same time) in a scene at >70 FPS (using my LOD system along with my polygon reduction algorithm) with nearly no visible quality loss.
fast dynamic shadows
Maybe it's just me, but I have never seen shadows, commercial or otherwise, that have not slowed down the game very considerably. If you want shadows, buy sswift's shadow system. It's fast and works well.
Another problem with working with blitz is that many more experienced people are aware of blitz limitations and are quite willing to ignore obvious problems in your work because they understand that you had to do pretty clever optimisations and used innovative techniques to overcome them.
In my experience, I have never run across a difficult barrier like you describe except for:
1. Shadows
2. Pixel shaders.
Both of these are unsupported by Blitz3D, and you are correct in regard to these. Although shadows are possible, the one reason one might choose not to use Blitz3D is it's lack of pixel shaders. Hopefully Max3D will be able to do all that Blitz3D can't.
And are incredulous when something like intercollision between animated objects doesn't really exist.
I have never had this problem, simply because when one of my games requires collision this complex, I have already decided to use one of the freely availible physics engines for Blitz3D. However, you
are correct that you would encounter problems with an
animated character if you require collisions for each limb.
Use the same textures throughout your levels and combine same brush meshes to lower surface count at the expense of propper culling of your scene.
I also agree with that. If you're going to make yourself a scene occlusion system for large indoor environments, the surface count will be higher than you could acheive with C++. No doubt Blitz3D does have it's limitations, although I simply disagree that it absolutely could not be used to produce a commercial quality game.
When Max3D is released, I hope it will provide all the "features" Blitz3D lacked. Naturally, I also would like Max3D to be released soon, but I wouldn't complain about it's release date. Of course, the sooner it can be released, the better; since some people have projects that can't seem to wait and move off to other engines, since they have no other choice (such as Red Ocktober who has switched over to DBPro). But it will be done when it's done, and if BRL rushes the release of it, it will most likely contain bugs or be unstable in some way.