It would be almost impossible to randomly assign those 192 settings to experimental units. Furthermore, once experiments are complete, experimenters tend to analyse the results as if the data were taken from completely randomised designs, allowing the results to be misinterpreted and ultimately slowing down the development process.