Simply solved by making the mechanical choices almost nonexistent a la 1e and letting in-game roleplay define the character.
Mechanical choices for my little guy in Monopoly are 'practically nonexistant', but that doesn't make it a good roleplaying game. You
can roleplay in Monopoly, but the scope is really limited.
And [a character's ability to interact with the world] is driven by in-game personality and-or decision making.
And the decision making has to be based on an understanding of how the world and the character's abilities work. That understanding can be really efficiently delivered to the player by the rules and systems.
Don't get me wrong - rules and mechanics are necessary to allow for a playable game - but they needs must get out of the way whenever possible. It sounds like (and please correct me if I'm wrong) you see the rules as the world; where I see the DM as the world, aided and abetted by the rules when needed.
The "DM as the world" is a broken concept. It can only be true of an author, and even then the reader supplies a significant portion of the world as soon as the book is actually read, rather than simply existing. In fact, I think perhaps daydreams may be the only times when lone "world creation" can be entirely the situation.
In a roleplaying game, the "reality" of the imagined world is shared - or else it has no meaning as a "game" at all. All of the players - including but not limited to the GM - therefore deliver input to the "shared imagined scenario". The rules, therefore, do far, far more than "aid and abet the DM" - they form the shared language, based upon which these contributions can build the shared "reality".
Without written rules you have the GM trying to convey the entirety of the world as they envision it to the assembled players - each of whom undoubtedly has a selection of preconceived ideas about the world generated by the first descriptions and what those descriptions trigger based on their earlier experience. This cannot happen perfectly in a 300+ page book; expecting it to happen in a few hours of conversation is, at best, wildly optimistic.
So, what generally results is (i) players relying on how well they know the GM to guess what picture of the game world the GM holds in his or her head, (ii) players having to repeatedly revise their own vision of the game world as it becomes clear that it does not accord with what the GM was thinking all along, (iii) some players doing better than others in the game, because their own world models more closely align with the GM's world model than other players, (iv) some other players frustrated because they need constantly to adjust their world model to fit a concept that makes limited sense to them, and a variety of other effects.
Rules and mechanics are a shorthand way to communicate how the game world works so that all the participants have a clear and shared view of that. This makes building a shared fantasy story immeasurably easier.
Absolutely disagree. The first is roleplaying. The second is game-playing; and role-playing - as in taking on a role and playing it - is pretty much nowhere to be seen.
Absolutely disagree
. The difference between the two has nothing to do with "roleplaying" by the definition used by you in the earlier parts of your post! "Roleplaying" has to do with making decisions and portraying the personality of the character - both of the examples given do that. Now, one of them does it with more flavour and from a more in-game-world point of view, it's true - but both display the character's personality and represent an in-character decision.
If you want game-world focussed flavour (and I take it from your comments here that you do), then I suggest taking a look at
why the player might, quite rationally and not out of some (lazily) assumed fault like "laziness", choose to announce a "use of Intimidate" and not a more colourful description of the character's actions. And I suggest that this relates directly to the "rules" and "DM fiat", curiously enough. If a player has a character who is good at Intimidation, but not good at, say Bluff, and the skill used is to be decided by the DM based on (the DM's conception of) the character's actions, then I can quite understand the player saying "Intimidate" quite explicitly as part of their action description. Saying "I use my strength to crush a mug, hoping to make the little rat fear me" leaves the door open for the DM to (honestly and without malice, perhaps) see this as a use of "Bluff" rather than "Intimidate" - and the player and character both are hosed.
The way to get more game-world focus and flavour into player action declaration, it seems to me, is to make it clear that the control over which skill is to be used lies with the player, not with the DM...
...unless they get lazy, and just declare a skill name alone, perhaps?
I think we agree that the game world has to operate with a sound internal consistency. Whether that comes from mechanics, DM fiat, player-driven input, or some combination of these and-or other things, is up to both the group and the game system.
Agreed - but there is a huge, steaming issue with it coming from DM fiat, which is that
the players need to know what it consists of! The only way I know of to address this is via communication; the rules are one very good medium for that communication. Another is a voluminous world description, possibly contained in a vast agglomeration of game supplements, world books, monster descriptions and other publications created over a period of (say, just as an example) 30 years or so.
Which is one reason why people who used this model for their game world communication were POed about 4e, of course. Personally, however, I have for some while used Hârn for this style of "world basis sharing" because it has, in my opinion, done a better job of it (more world supplements than rules, fixed time point for all description, etc.).
The way to beat the optimzers is to make the mechanics simple enough that there is really nothing to optimize.
Failing that, the only other way to blunt them a bit is to put something loud and clear in the PH about playing to the spirit of the game being hugely preferable over trying to break the game, and then hope each group sees fit to run the optimizers out.
Why? Why declare one mode of fun "badwrongfun" and deny those who enjoy it any place in the game? That strikes me as just bigoted and spiteful.
On the "pre-gens are always suboptimal", for example, it strikes me that an open playtest offers an ideal way to approach this. Let the charop guys play with the self-build system for a few months; you'll soon find which are the optimal builds. If the builds they generate don't give good flavour, tweak the "feats" and suchlike elements until they do. Then use the optimised builds as your "Themes". Eliminate or buff up the choices not included in the Themes at the same time so that those who want a different flavour to their character have options that, while not as good as the "standard" characters are not disastrous, and you're done.
Oh - and then don't succumb to the temptation to publish a slew of new feats that break the optimal nature of the original Themes to make a fast buck. That might be the tough part, actually...
. With a caveat, of course. I think a singular class is a little too restrictive, but I think feats and skill points were a little too loose (and just killed DM prep time.) So, I kinda like backgrounds and themes for the middle ground they provide. They give you a decent amount of variety, so that a 3-fighter party can actually have three unique characters. Really, that's all I want from a customization system. 
