Isn't the whole idea McDowell not buying Hume's 'direction of fit'? I agree that splitting things to re-merge them is a bit goofy, when the author contends the splitting is artificial.
But to me this is just vocabulary, the point is that there is a single, undivided moral 'fit'. Since the language traditionally creates a false division, you can describe it as a merger of the two more ingrained concepts. So it is not a 'new' mental state, it is the original 'real' mental state that has been artificially misinterpreted by Humeans as a blend of two states that just seem easier to express.
On the topic in general I believe there is more detail rather than less, and I reject McDowell's move toward holism here. Beliefs and desires are separate things. They may be the endpoints of a gamut, but the distinction is not misleading.
I still agree that this reciprocal mind-to-world/world-to-mind fit is inappropriate. Hume's pair needs more structure, not less. It leaves out the way that the individual world interacts with the rest of human thought. I much prefer the simplification of Lacan that reflects a traditional Alchemical Cardinal/Fixed/Mutable division.
In all ways, but morally in particular, we live in three worlds, the physical, the imaginal, and the symbolic. There is no fit of mind directly to world, but only one mediated by that mind attempting to negotiate its place in the social structure, and there is no direct fit of world to mind except through that mind projecting concepts derived from the social structure upon that world. So Lacan's Symbolic Realm mediates all of our attempts at such fitting together.
As a direct answer to your question, then, no: to a large degree, our moral development is coping via cycling more than fitting via dynamic tension. Actions have effects that are perceived and extrapolated by the imagination and attributed to symbols which evoke emotions that motivate actions. (Rinse, by repeating?)
Belief is a symbol as a clear image. Desire is a symbol as an emotion (a physiological effect.) But there is a symbol there to back either of them up, the moral motivation is the need to have a working set of symbols to base your psychology. Psychological homeostasis is motivationally "ert". It is highly energized and has a definite motive.
(Perhaps off topic:
It is tempting to merge the imaginal and symbolic worlds, but if you do so, there is no mind. You have to distribute the processing we think of as done by a mind out into the culture in general, and that seems bizarre and messy. It seems more tractable, and truer to our actual feelings to think of a larger group-mind that works in a significantly different way than our individual mind, but which both contains and pervades our individual thinking.
That way instead of there being no mind, there are layers of mind, with a purely imaginal, purely individual mind at the core, and layers of group minds with different levels of symbolic investment around it, at some point becoming relatively static symbolic structures that do not think.
For example, the cultural value of empathy is a thing, not a bunch of coordinated separate things in different minds. It derives from a common cause well before any currently existing human mind, (as do its opponents like skepticism, economics, and defensiveness). And actual judgments of the appropriate level of empathy are negotiated with that thing, not fitted into our lives piecemeal according to our whims.
Now hopefully back on topic:
The rule-following part of morality arises because the symbol functions and connects with the minds that use it by becoming embedded in real life as a 'game' in Wittgenstein's sense. Actions are actions, and whether or not they are moral is reflected by their interpretation. The 'rightness' of a move in a language game is an emotion, which has an image, which has an interpretation, which refines the move, or the game.
The game captures desires or concretizes beliefs in statements or actions and feeds back images of the effects of the symbol to itself. A common moral sentiment leads us to constantly try to capture our language-games in rulesets, even though they cannot be properly expressed in this way. And we end up following rules.