I'm not sure even this is true. The city may not build up in the way the planners anticipated. Even if it does, it will change over time.
You build your infrastructure to, or just past the edges of what you know is being built, and make sure you reserve space past that to accompany further expansion.
The lets you expand the ~factory~ city while still allowing some sanity to the underlying infrastructure.
* This will likely apply to anything that needs both preplanning and unknown future expansion such as codebases
Bus lines instead tend to follow where the traffic wants to go. Around here, many shopping malls double as bus stations because the primary aim of transit seems to be circulating consumers around places they will spend money.
This is circular reasoning though. Bus lines go to the mall, but malls are built where the bus lines go. Malls use a lot of space, so carving out a little bit for transit is easy.
Trying to avoid painting yourself into a corner without over thinking is the game.
In the cases I know of the density followed upon building the line, but that may not be universally true.
It might not, but planners have a lot of forcing functions.
For example, if they build a transformational backbone, and nobody uses some part of it, they can fix it, by say, moving major government offices to that area. Residents, other businesses, services will follow suit.
What doesn't work is half-assed planning, where you have a grand, but myopic vision, implement it, and never adjust after-the-fact.
The out-of-address-space v4 that wasn't future-proof enough to conceive of more than 4 billion computers?
Or the reinvented ipv6 that seems to have gone way too far in the other direction (and not just in terms of address size)?