I especially love the problem space of transit maps. Transit feels like a purely topological endeavour. What nodes connect, how, and what are the edge lengths? But the moment you want to act outside the transit system, you then care about Cartesian accuracy as well. You want to know how far something is from two different transit stops. These kinds of maps can really trick you big time.
When it comes to static maps, you’ve gotta make a compromise. But when it comes to dynamic maps, we could do a ton more. Imagine a map that can trivially toggle and animate between being transit system legible, and surface-level legible. Where in one mode all the lines are nicely laid out to see how they link and their trip times. And another mode every node animates to a new position where they represent the Cartesian coordinates on the surface (another lie of course.)