I don’t. First part of the anwer is a mix of polygonal modeling basics and knowing about normal mapping. We’re dealing with approximation for visual purposes and the accuracy is relative instead of absolute. The accuracy both in terms of dimensions and detail decreases when the distance increases is how the human eye works, so it’s exploited everywhere: games, movies, drawing, painting, physical and digital mediums.
Normal mapping doesn’t affect the geometry so it can’t affect the silhouette, so when two elements overlap it can break the illusion. If the camera is close enough or the detail is big enough, it will be apparent it’s not real.
Retopology is just the name for the technique I suggested. Topology is a term that can mean a huge amount of aspects related to model structures, but simplified it means the structure of the model and edge/face flows in it. When we have a form with some structure and we build another structure for the same form, it’s retopology.
In practice it means that the new structure is built on top of the existing form, so we’re creating a new structure but don’t need to rebuild the form. There are several ways to conform the new geometry to the old one, shrinkwrap modifier is one of them.
Modifiers you learn by reading the manual and simply trying. Some are less used and more involved, which require a tutorial or even several tutorials.