Where it says "the iron ore is then directed to go to the city of Torgeto" in camp #1, this is a "push" model. This is made more apparent by the mention of the sliders and ratios directing how much of each resource gets sent where. We are directing the resources from their point of origin. To do this without automation is potentially a huge pain; instead of clicking a central point and directing things to filter into it, you have to go round every exterior point that needs to be redirected and readjust it manually. Readjusting things on a "pull" model, like camp #3 - even without explicit automation - can be done from that one central point, and this is true even if you forced the user to create each route from an outlying area individually.
Yeah, I guess there is a difference. Nonetheless it seems to be pretty non-fundamental. The overall economy in the game would be quite similar either way, the 'push' vs. 'pull' would mostly just affect the automation system. I would hope that whatever override mechanisms are in place will allow the player to make commands in the most relevant possible way. If I want to order all my iron ore from all my iron mines to be shipped to one city, I would like to be able to effect that from the city. On the other hand, if I want to split my iron distribution from one iron mine between several cities (in some way other than is done automatically) then I want to be able to do so at the mine (or in some management interface, which ideally would be constructed to allow you to make changes via both methods).
The pull method is fundamentally more suited to automation than a push method, but there is no reason why one should be favored over the other when it comes to direct player control.
There are a number of issues with AI governors, which are a different thing again. Unless what they do is immediately apparent to the player, there's a strong risk of them making decisions that the player wouldn't in their place and doesn't know they're going to make. This doesn't necessarily make them bad decisions, but the player either has to accept a niggling doubt as to what the AI governor is doing, or spend time watching it to see what it's up to, both of which lead to unsatisfying gameplay.
In short, "automation" is not a magic wand that will solve all user interface issues, and thought needs to be given as to exactly what form such automation ought to take and what the interface should involve. Some of the details of that interface are already being discussed (sliders/ratios in camp #1, "pulling" resources in camp #3), and deciding on these details will constrain what automation possibilities are available.
Well yeah. Automation will always be suboptimal, at least until we can plug our computers into our heads. Without being aware of our thoughts, the computer is forced to make strategic decisions based on an incomplete subset of relevant information. That said, I think that in most cases it can be done well enough with regards to the economy in the vast majority of situations. I think that a computer is actually much better suited to the task of handling a complete economy than most people, unless someone is willing to literally spend 20 minutes a turn with a spreadsheet in hand and carefully planning out the next several turns. In other words, I actually think that this might be one situation where, at the large scale, the computer would be better than the player. The player only really becomes important when something goes wrong, or something comes up that the computer isn't in a position to recognize (like a long-term player strategy). And in those cases, it's likely that the player only cares about a very small subset of the economy.
That said, of course automation is not a magic wand that will solve all the problems in the world. But I think that this is a situation where automation can truly shine (unlike, for example, Civ IV workers, who are consistently stupid with occasional bouts of intelligence). I think that in the case of a sophisticated economy model, it would be very feasible to create an automation system that would be consistently intelligent with occasional bouts of stupidity (or lack of foresight, or lack of mind-reading ability ). It's a problem uniquely suited to automation (if done well, obviously - and hence this prolonged discussion about it). The level of player intervention would vary from person to person. Some people might be happy to let the automation do its thing except in urgent situations, while others might constantly tweak things here and there each turn to make things just a little bit more efficient. But I doubt anyone would be seriously tempted to turn off the automation, whip out a spreadsheet and take complete control... That would just suck.
If you can't design such interface, then your entire argument falls apart. And you know, you can't do it.
Saying, "it can't be done" simply indicates a lack of imagination. It would be difficult, for sure, but Stardock's specialty is UI creation. To display all of the information about all the distribution and management of every type of resource all at once would be foolish - it would be overwhelming and convoluted and downright unhelpful. So break it down into subsections, display the most important aspects of it on the overland map (maybe at the toggle of a key). Make it able to display your production lines. Chances are no one would be actively using all 110 resources at once, thus reducing the number they'd have to care about; the UI should take this into consideration and hide information about resources that aren't being used (this is more for MR than NR). If a player is not using spears, and is not making spears, then the resource management interface shouldn't be cluttered with information about spears.