Probably the one vector that work does have from the ivory tower to the real world is via people, when they transfer from one to the other, or have one foot in each side, and can transfer ideas and PDFs from the ivory tower to the real world, and can transfer motivations for new ideas from the real world to the ivory tower. Bridge-people like these are some of the most admirable people in academia, in my opinion, and also in my opinion produce by far the best papers.
(I am therefore trying to be one of them, sort of, although I'm not really interested in actually writing papers, since to get it published (and therefore have it count officially as a 'paper' instead of a PDF talking about what you did) you have to write a variable amount of obfuscatory bullshit.)
I agree there is something wrong with the model, but I think your understanding of academia is flawed.
Which is sad, because the individuals publish there in order to move up through the academic ranks. By moving forward in their careers, they are also marginalizing their own work.
The incentives are perverse, and it will be a very happy day when all CS content is freely available online, as it should be.