The simple answer is that everything can be done without raytracing but using raytracing is easier to implement and will in the end give better results and often in a shorter time (given the same accuracy).
If your research leads you to another conclusion you are of course free to build a renderer according to your research. If you build the perfect renderer (accurate and fast) I am sure you will find a market for it.
In the opening message you asked for our opinion, we have provided this. If you are not pleased with the answers seek answers elsewhere. This discussion is not leading anywhere if you don't want to hear about the implementations using raytracing. Once you know what has been implemented using this technique and you have studied those implementations, you can draw your own conclusions and implement whatever you like best.