Maybe it's just me, but I find the standard pattern of gaining XP by making an object, and then using that XP to gain a skill level in making objects, lethally boring.
The only thing this conception of "crafting" can possibly lead to is grinding. Crafting becomes dominated by Achievers who are willing to grind a zillion worthless objects in order to gain the title of "Master Whatever."
Blecch. As an Explorer, why should I pay to play a game like that?
I think the problem with this "XP from making stuff" paradigm in RPGs is that it shifts the learning process to the character... but it's the player for whom learning is fun! When the crafting game is designed so that it's the character who "learns" (i.e., gains a skill level by making stuff), that form of entertainment is taken away from the Explorer players who would enjoy it.
Role-Playing Games allow you-as-a-character to do things that you-as-a-real-person can't do. So it makes sense to design an RPG to allow characters to have increasible skills in things like Broadsword and Fireball and so on. But it's a mistake to apply that approach blindly to everything -- if it's something that players can reasonably be expected to do (and enjoy doing), and it's OK in the context of the game for them to do it, then it shouldn't be abstracted as a character skill.
Crafting as a form of discovery is, I think, one of those things that players can and should do themselves. It's the player's ability to understand complex systems that should determine the character's effectiveness in the game -- a grindable system for mimicking this ability only takes fun out of the game for these players.
Accordingly, crafting as I'd like to see it designed into a game (not necessarily every game, but some game somewhere) would do two things.
First, it would recognize that different kinds of players have different reasons for wanting to make things -- in particular, that there are "discovery crafters" and "sales crafters." And it would therefore offer two separate but related crafting disciplines: Research and Manufacturing. (The actual discipline names would be picked to reflect the period and genre.) Research would be about letting Explorers learn how to make new things; Manufacturing would enable Achievers to mass-produce items to compete in the sales game.
Second, the Research game would shift discovery-based learning away from the character and over to the player. The fun would be in letting the player figure out the rules of the system through thoughtful experimentation. (No, of course other players wouldn't consider this fun, but they've already got plenty of content focused on their entertainment style.)
Gameplay would involve using knowledge about the properties and behaviors of materials to collect more such knowledge. Players would devise tests for the player's Researcher character to run that apply selected materials to each other in new ways to see what happens. (In other words, players would attempt to discover processes by which the properties or forms of materials can be changed.) The character would then do the scutwork of running those tests (possibly with some +/- modification of the results based on the character's "skill" level), and the player would be presented with the results.
If the results aren't sufficiently better than some existing process (and they typically shouldn't be), enough information will be returned so that the player can think of another test to try. This is the "learning from mistakes" part of discovery-crafting. (Trying to figure out how to let characters in an RPG learn from failures is unnecessary and undesirable -- it's the player whose imagination is being challenged who needs to be rewarded.)
If the results of an experiment clearly show that a better way to do something has been identified, then the player has found a new process. Once it's properly defined and documented, this new process can then be made available (through patent licensing to any/all Manufacturing players or outright sale to an individual Manufacturing player) in exchange for resources (money, materials, labs, directorships, etc.) that allow for more advanced experimentation. (Note: The details of this interface would have to be worked out carefully to insure that Achievers aren't motivated to create Researcher alts that would crowd out actual Explorer crafters.)
The virtue of this dual-path approach to designing crafting is that it doesn't confuse what Explorers want with what Achievers want. By separating Research from Manufacturing, Achievers retain the crafting-as-sales game that they're accustomed to and expect while Explorers finally get a discovery-centric game feature that's worth the money they pay in subscription fees. And by carefully designing these two kinds of crafting to work together to produce desirable goods, all the other players of the game benefit.
In summary, crafting should be about more than just the mass production of identical items. It can and should also reward the distinctive creativity of individual players.
I'd really like to play that game.