I posted both comments, I thought he was talking about the buses.
If you leave it to a typical topology and scheduling optimization process (with both automated and manual reasoning), after a lot of effort it would result in lots of small or big changes to the tracks and possible schedules to follow to maximize everything you need (passenger and freight).
For example, many times you don't need entirely new tracks, only a way to avoid trains conflicting in the same track due to collision or slow speed (a new path in existing tracks to follow, small parallel tracks in certain key locations, ...).
Signalling and geometry can be improved with technology, showing which speed the driver should use in each segment of the trip, where are the other trains, his whole trip plan in detail, changing it in real time to reduce trip time or avoid accidents, ...
Changing speeds also help with the geometry if you can't afford to change it.
Some trains tried to minimize the problems with geometry by leaning, it didn't work too well, but if somebody finds a way to shift the train's weight without tilting the passenger's cabin, then it's also a possible idea.