I'm a huge advocate of the rail-road split, but "instantaneous" and "unlimited" seem like taking it too far. Rails move "lots of stuff" "quickly" but supplies going from Berlin to Kiev are still going to take longer than a day, even with a good rail connection.
It sure does. An ideal system would not be instantaneous or unlimited. But the capacity of a rail line is very large, and along most operational rail tracks, all major powers of the war were more or less capable of supplying at least the ammunition and food that was needed (possibly with the exception of things like artillery shells).
Human planners can be relatively capable of planning basic provisions in such a way that the soldiers will not run out of food or bullets too quickly, as long as the total throughput is sufficient. For heavier equipment, countries like the USA and the UK did face some bottlenecks in their capability to transfer equipment overseas, but for land based nations (such as Germany and the USSR), the main challenges was to move their equipment from their rail heads to the soldiers at the front, especially when the distance from functional rail to the front was large. (Germany also had big issues with producing sufficient amounts of ammunition, food and fuel in the late war, and had to save up a stockpile in order to conduct operations such as the battle of the Bulge, but this was a limit of production capacity more than logistics).
Now, if you could make a computer AI that would do all the logistics from the factory to the rail head, it would be neat, but from my understanding, does not add so much realism (except for overseas supply). Instead, the effort put into the supply system should be focused on imposing the historical constraints that units faced during the war when operating too far from their rail head (for overland supply), as well on getting the capacity of overseas supply "just right".
Based on the assumption that transporting supplies along railroads was not the bottleneck during the war, I suggest that this aspect is removed from the game in order to make a system that is simpler for the developers to optimize in a predictable way, even for countries/coalitions that control a lot of territory.
Maybe there are other ways to solve this than my suggestion, but I have a feeling that there may be difficult to make a system with finite throughput and speed of supplies that will:
1) Avoid artificial bottlenecks around the capital (or other primary supply source)
2) Avoid similar bottlenecks in areas where a lot of supplies are transported, but where the aggregate capacity should be plentiful (such as Poland).
3) Provide a reliable way for the supply AI to judge if supply should be sent over land or if an overseas supply route should be set up, as well as how to set up automatic supply depots.
4) Prevent the kind of oscillation effects seen in HoI 3, where supply is sent back and forth between units and the capital all the time once supply capacity starts to be constrained.
Maybe player/AI controlled forward depots can help a bit, as it makes the movement of supplies to the combat divisions more of a local thing, but moving supplies to the forward depot could still be a problem. With unlimited speed/capacity along railroads, I think the need for player controlled depots is much smaller, as any territory with a rail connection to the central stockpile can be treated as if it were a local depot.
By making the central part of the logistics chain easier, the developer can focus on making the local supply situation much tougher (and more realistic), especially when in combat. Even a panzer division, which carried its own supplies, could only carry enough for 3-4 days of combat. In HOI 3, units would bring along 30 days of supplies and fuel in order, so that the problems with the supply system would be suppressed a bit. As we know, this led to a lot of other problems.