I would recommend as least something based on LttP or Minish Cap(which somewhat combines GB and LttP mechanics). However, my suggestion would be the onion approach. Go through a layer at a time starting with the general stuff then slowly working to the core components of what is in the engine such as specific items, ai, and so on.
The modular approach (or 'onion,' as you put it..) is what I generally aim for with my tutorials. Each tutorial will add a new piece to the puzzle until the information in the tutorial can produce an entire game. This structure makes it a lot easier for users looking for something outside of the complete project to come in and figure out something specific to another task they are looking at (such as initializing an OpenGL context with SDL, or implementing the backend for an inventory system).
Another thing to think about is editability(not a real word but only what I can think of with 3 hours of sleep in 48 hours). What I mean by that, is how accessible would the engine be on all levels of skill? You are obviously going to have everyone from experienced programmers to noobs going through each. They may want to do different things with the engine/tutorial than you might have envisioned. The more experienced could expand on what you come up with and create something more crisp or in greater thought in certain areas whereas the noobs may just want to know how to make an "original" item or change inventory layouts. Hope you understand what I am trying to say. Just a thought...
The level of the code I use in this project will be determined by the general desire from the community. The tutorial series is going to walk through the programming of the entire engine, from creating the project in Xcode/Visual Studio/MonoDevelop to the code used to import the game maps. Regardless of the difficulty level within the tutorial series, I do intend to have game content created with external tools - tools such as Tiled and those that will be developed specifically to support the tutorial series. Each of the tools will be open-source, though I don't plan to cover the tool creation in
this tutorial series (though I may reserve that for later tutorials).
However, the level of difficulty desired by the community will dictate how static or dynamic things such as inventory layouts and inventory mechanics are for less experienced developers. Taking your "original item" example, readers could examine the methods for creating several different items - looking at the factors that remain the same between all objects (selecting the icon, naming, tying them into the game, etc), and looking at factors that are similar with regard to mechanics in a smaller subset of the items (such as creating throwable items).
Also, I wouldn't mind some GLSL. I am still trying to figure out what exactly I can do with HLSL so it could give someone like me some insight into that as well.
The general ideas remain the same between both - for example, what Fragment shaders (pixel shaders in HLSL/DirectX) and Vertex shaders each handle remains the same. The difference is one of semantics and implementation (GLSL doesn't, for example, use the pass and technique system that HLSL aficionados are used to).
In the end to your last bullet, you could go the route with implementations of elements unique to each game in small amounts: direct scrolling(LoZ, GB), stat/leveling system(AoL), mode 7 map(LttP), mini quest collection like the friend toekns or whatever they are from Minish Cap, or even multiplayer like the Four Swords.
Not saying to go very in depth but just some basic implementations or even crossovers.
I wasn't necessarily referring to individual elements unique to the games, but the overall feel of the game. For example, if I spend the tutorial series going over developing a game styled after the GameBoy games, it will be a vastly different experience from the GBA/SNES games' style. As far as individual elements for a game (such as side quests & distractions) are concerned, I would love to hear suggestions from anybody as to what specifically the community would like to see. Some elements, such as creating a "mode 7" map (which is actually both a great idea, and a
very simple thing to implement) are easy implementations that can be done with a single tutorial release, while others (such as networked multiplayer support) are things that must be considered very early on in order to be even feasible, and even then will have drastic and very time-consuming implementation tutorials.
More than likely, I will not include multiplayer features along the lines of Four Swords. Given interest, however, multiplayer mini-games such as PvP arenas are not entirely out of the question - they may not have much of a Zelda feel, but they would be more than sufficient to cover networking fundamentals without bogging down the users who either do not want to look at such high-level tutorials, or would not be able to follow them.
It's important to note that I'm not seeking to create an exact replica of any individual game's features - that was mostly a question of what overarching play style the community would like to see.
Personally, I would like to see a Zelda tutorial series that contains the original elements but adds things like dynamic lighting or visual effects instead of everyone usually resorting to black overlays/cloud overlays/etc.
This is exactly the kind of suggestions I was looking for, and I completely agree with you. In the tutorial series, I intend to cover a variety of special effects systems, including particle systems, shaders, overlays, and sequenced effects.