With a few quirks, you described how Hotspot works. On a more meta level, you should also account for the quantity and the quality of the data that your constraint solver has to work with. To make a parallel, think on how query optimizers in SQL engines change the execution plan based on collected insights on data stored.