when it's impossible to make good looking games with it
Well of course (rolls eyes).
From my point of view, Blitz3d is nice and simple and flexible to use. There's a number of powerful libs that extend its capabilities (ParticleCandy, SpriteCandy, shadow libs, pathfinding code etc..) and I have an extensive codebase to work from.
It's also stable - which is not something to be overlooked. When your game finally hits the public, having a forum full of people angrily complaining about crashes/freezes is neither desirable nor fun. I suspect that the people who put down B3d for not having the latest shader fx don't take enduser stability into much account. I've had a few bugs pop up with TU but the worst of these (Vista crash), was actually down to a programming error - not Blitz.
Having said that, Blitz is not very fast - both in running code and the graphics engine.
Which is a quandary - if I go to Blitzmax I lose all of the excellent libs much of my codebase depends upon - and my codebase would need to be rewritten.
If I stay with Blitz3d but use a new engine (via dll), I'd still have a lot of recoding to look at. So from my point of view, Blitz3d2.0 would be the best solution. The core speed of blitz may not be improved, but the graphics engine would likely be faster. Mark mentioned something may be available around Chrissy next - I certainly hope that would be the case.