...are the ones who can really exploit all three possibilities. In 2001, a record test train of almost 100,000 tons (in it 82,262 tons iron ore) ran on BHP Billiton's Pilbara line in Australia. The eight 6000-HP locomotives pulled 682 cars - 7.4 km in length! But even regular trains are up to half that long. For comparison with the next two: axleload is now up to 40 t.

USA

US railroads, with not much to limit their size, and a network of long lines relatively simple to upgrade, also grew rather big. But, while high axleloads (30-35 metric tons) would make stuffing two containers upon each other only logical, doing so in practice was a formidable challenge even in the USA (it started only in 1977).

The center of gravity had to be lowered somehow (wind is a danger), and even with low-floor cars, the double-stack loading gauge, a metre higher than normal, forced some tunnel/overpass rebuilding - in Europe, they'd tear down overhead wires¹...

Europe (-Russia)

On our continent, a network too complex and with lots of stations won't allow an increase of train length. Dense buildup, historical buildings and 'well-built' old railroads mean loading gauge can't be increased. Unless a lot of track is replaced, raising axleload won't be practical either - so just on mainlines, it took half a century to go from 15-20 tons to today's 20-25 tons.

But, some tricks can still be made. Two decades ago, to transport the biggest (=highest) lorries, engineers designed special low-floor-throughout RoLa² cars: with wheels a third of a normal freight car's in diameter, and four-five wheelsets on a bogie instead of two (so that axles don't break).

(I note that where I work, my collagues once had to investigate these cars - they proved that their derailment conditions are completely different from normal wagons, f.e. speeding across switches decreases the risk!)