filtering – FFT Windows – Looking for high-resolution, low sidelobes

So I am pretty familiar with FFTs and FFT windows. I know that in general there is a tradeoff between mainlobe width and sidelobe levels.

However, I recently stumbled across this discussion, where someone claims to have created a window or FFT method that achieves both very good mainlobe resolution AND very low sidelobes. However, he doesn’t explain the technique.

Looking at the frequency response, I get the impression that there is maybe some upsampling/autocorrelation involved, or maybe some iterative technique.

Does anyone know how this might be achieved?