Errikos is right on the point with this discussion: we are not striving for utopia, we are striving for an improvement, a better relationship between what the composer inputs on the score and the samples available. I speak for myself, of course, but I can assume such was the feelings of several VSL users with the release of Sibelius 7. There was no improvement to integration of VSTs, and I dare to say, no improvements AT ALL.
Some of my burden with the available notation software is precisely the focus which is placed on the publishing technology, I believe this is part of the expectation we face, that the composition must be brought from elsewhere (DAW). I would be happy with an abstract notation editor, where only the composition is in focus, with absolutely no need for page layouts, but I understand some people are more comfortable having the layout in place, which I believe is more similar to the natural notation on paper.
If VSL would take the helm and develop a notation software based on its libraries, of course, the result would be spectacular. If this is not possible, I can only hope another company will take the task of developing notation FOR orchestral samples, or that at least, the available notation packages allow the user to interact more friendly with the midi-data.
The problem with notation playback as it is now, is the extra step that midi information takes to access a sample. As said before in this thread, in DAWs the user communicates the midi data directly with the VST, while in Notation the user places symbols, which in turn are converted to midi-data and then reach the VST. These symbols, I believe, are the biggest difference between Notation composers and DAW composers. DAW composers want to input data, numbers or graphs, which will affect the sounds being played. Notation composers want to input Symbols and expect them to generate the data accurately.
Notion advanced this a little bit, allowing the user to have a set of interpretation rules per document, per score, which means for a particular performance, you can adjust your rule files without interfering with all rules for the VST, which is what a soundset does. Still, this is far from being able to adjust one specific instance of a symbol in the score, like dshertz said, what's the difference between a sfz and an altered sfz? I believe this is similar to what Sibelius Properties window achieve.
In my imagination, the easiest way to provide better performance for notation would be adding a control lane below each staff, collapsable of course. Or better yet, a secondary type of notation that could send changes for the dictionary itself and not to the VST. Using the sfz example from above, let's say the dictionary has the sfz as an +10 increase in velocity. You could apply an extra symbol to the staff, not a number or a graphic, but a different symbol, that extra type of notation specifically for adjusting notation playback itself, maybe on different collor. These would not print or affect the layout of the score, but would affect playback, they would mimic specific decisions the live performer would have made.
Exploring this idea a little further, in the same example above, let's say the composer wants to lessen the sfz a bit, maybe by -2. He adds the -2 Velocity symbol, or maybe, even add it twice. For creating a drag in tempo, for instance, instead of adding a sequence of tempo changes to replace an abstract text instruction, he could add an inverted tempo hairpin, that drags the tempo precisely where he wants, just like a crescendo, he only places the starting tempo and the ending one.
Thus he achieves most actions through a type of notation, which saves the composer the hassle of dealing with yet another type of information: midi. Of course this only matters if the symbols are preconfigured, designed to take the place of every action that is commonly used in midi finetuning. The point is, every change to the performance is done through notation, even if of a different kind.