There was a short period of time where software rendering engines were able to do some interesting things before hardware rendering took over and they all died out. There is still untapped potential in a lot of the techniques, which is fun exploring.
I'm using them for a game I will be releasing. I think very few games these days regardless of engine used or graphical style has a chance of a net positive return, so that part doesn't bother me.
Yeah, it's neat to see how people solved problems. It feels good to say "this is the algorithm that adds lighting" rather than "I set the camera at xyz, a light at xyz, a model at xyz, and called render()". Today I added some code to create comic-style outlines of objects.
I have a small library of books from the era on 3D graphics that are fun to read through (I especially like the speculation on how DOOM achieved its speed and "realism"). Andre LaMothe and Christopher Lampton have some good stuff, plus the black book of Michael Abrash.
I tried doing indie gamedev as a full time job for a year a while back, learned that was a quick way to lose all my savings :) so now I do it as a side project while having a full time job. If any of my games bring in some pocket change, all the better. Less stress for sure but also much slower progress. My brother does the music and sounds, and I do the code and graphics (plus some commissioned assets).
I like this as well, so my guess is that some people grew with an interest in game development, and they still find it interesting to explore it.
Game algorithms were also "not easy" in the 90s, but since the codebases were overall smaller, they're more approachable then modern ones.