You state: "I always wonder whether those who argue for no separation think weak, inexperienced cyclists should just stay home or just ride around in circles in the schoolyard parking lots on the weekend." These are the cyclists who refuse to obey the rules of the road for drivers of vehicles (RRDV). Officially, the only cycling skill such people are expected to possess is riding a straight line. But that is not the average; most cyclists have a few traffic skills: stop signs, traffic lights, perhaps signalling before turning (which is not what makes turning safe). So now consider. In what way does separation, or type of separation, make cycling safe for such people? That is, what is the mechanism that greatly reduces car-bike collisions that would be incurred by such cyclists? Here's a big hint. You need to know the kinds and frequencies of the car-bike collisions incurred by these cyclists before you can make any evaluation of the safety of any road design.
VC cycling, generally defined as obeying the RRDV, is directed at greatly reducing the great proportion of the car-bike collisions incurred by the general cycling public. 80% is not a bad estimate. But those cyclists who refuse to obey the RRDV cannot benefit from doing so. For all their demands for safety, they are eliminating their best chance of getting safer cycling. Given that refusal, what is the traffic-designer to do about such cyclists? That's a damned difficult question.
One can, of course, largely ignore safety and simply do what these cyclists want. They think that it makes them safer, but it really has little effect. In fact, that's what gets done. So, just muddle along doing the best that you can to satisfy the general cycling public. But take care that what you do, or what the government associates with what you do, like anti-cyclist Far-to-the-Right laws and such, does not prohibit those cyclists who prefer riding safely according to the RRDV instead of doing what the superstitious public wants.