So then why bother with cities at all? High rises are just a different constant factor relative to 40 acre farms after all.
I feel like this is analogous to a case where someone says that an algorithm only differs by a constant factor but it turns out that because of that difference it hits the cache for 99% of use cases and as a result you see better than 100x speedups for all real world workloads.
Cities with cars obviously work up to some size. The sheer number of personal vehicle trips that a single rail line can replace is huge. And just to give you an idea of how large the scale factor here is, consider that you can pack at least 3 rail lines into the width of a typical two lane road, and that large cities commonly have 4 lane arterials.
If the political will existed to reallocate the space it could be done and the viable density would scale accordingly. On the extreme end we have Tokyo as a practical example, and even they are far from saturating all the available space for building rail lines.
> moving goods instead of people scales logarithmically
This is obviously false. There's some average parcel size, and an associated maximum capacity for a delivery van. Thus any given delivery run has a limit on the number of shipments it can service. Obviously that scale factor is significantly larger than the one for passenger rail versus cars, but it's still "just" a constant factor and thus irrelevant for scalability by your own logic.
In fact by your own logic I'm fairly certain that you will find that life as a whole is unscalable. Better nuke the planet I guess.