Interesting. I actually find the scaling to work quite well, in regards to population at least. Efficient subways are possible in simutrans, they just have to be built more like real subway systems.
1. The face of the map should be so built up that only a subway can navigate the landscape without excessive bridges/underpasses or deleting of public buildings (this on its own doesnt guarantee anything, but this is how subways work--they support an already robust network, providing an *overall* advantage on the map, not just to their own lines)
2. The services should be long. Someone earlier mentioned NYC has 400-something odd subway stations. This is true, but the 6 train, for example, has I think 38 stops on it. MTA in NYC has 22 separate services, with special express services on some of those normal services. This brings us to the next point
3. Systems should share as many stations as geographically convenient.
4. Population density needs to be extremely high over a large area. To use NYC as an example again, there are about 175 city blocks that make up the length of what most people think of as Manhattan. Imagine how big of a city that is in simutrans. If you were to build a city this size in the game, it could support the previously mentioned 6 train which spans 38 stops. We know this is true because it is easily done in ny911's New York scenario
5. You need to be well networked. Just as in real life, your networks need to be linked to airports, bus stops, and normal train stops to have the ridership to support their cost.
If we are talking about using subways as early game or aesthetic tools, then maybe changes should be made. But as it is, I think simutrans reflects nicely the purpose and risks and etc involved in a subway. When you consider that 1km of subway could easily cost 250 million dollars, or that right now, NYC's new "T" line is costing 1.7billion per kilometer, we should be thankful
Seems to me simutrans needs to have another discussion about realism and fun and how big the grey area between is