Earlier quoted context omitted.
There's no need to have only one line to connect everything, even in a linear structure. It's still about the topology. At the simplest there would be one line connecting from end to end, which should be the fastest and the one with the fewest stops. But then you have smaller lines, travelling in their own tubes, connecting local areas. That way you can travel between any two points relatively fast and avoid congesti…
In 2D, you can build a grid of straight line railways. Riders transfer zero or one times. Cost scales proportionally to area (and hence population) assuming constant density. Time between two random points scales as sqrt(size). Ridership on a given rail segment scales like sqrt(size), which may be tolerable. In 1D, you can build a linear railway. Cost scales proportionally to length (and hence population) if you assu…
The length is also much less of a concern than in traditional transport systems because of less friction and less need to brake, so most of the energy is for accelerating and decelerating at the stops.
Of course it would blow up if you wanted to build a fully connected city that is twice that length or even longer, but that's not proposed.