How to properly separate render thread from logic thread?

Question

I am not asking about parallelization of physics and so on stuff with parallel_for() like stuff.

I've managed to have separate rendering thread from logic thread, in glfw3 application:

void render_thread() {
    glfwMakeContextCurrent(win);
    while( !glfwWindowShouldClose(win) ) {
        sleep(10); // emulate heavy rendering, to be sure that main loop stops event processing during this sleep
        glfwSetWindowTitle(win, "some title"); // posts message? to window queue to update title
    }
    glfwMakeContextCurrent(nullptr);
}

And init/event/input handling loop is as follows:

void main_thread() {
    set_thread_priority(highest); // to be sure this is first priority in OS sheduler, compared to render thread
    win = glfwCreateWindow(...);
    std::thread rt(&render_thread);
    while (!glfwWindowShouldClose(_glWindow))
        glfwWaitEvents(); // could put thread to sleep
    rt.join();
}

Here is the problem, when i click to close window:

• rt.join() never returns, because it waits for render_thread() to return
• render_thread() could not return from glfwSetWindowTitle() call, seems no message processing goes on, or something other that i dont know.

Maybe my whole approach is wrong, but how should i solve that in a sane way?