An article by a digital musician I read recently claimed that although digital synthesis can approach the quality of a real orchestra, it's extremely time-consuming to shape every note to fit the mood and context of that note.
If you factor in the time and effort to "carve" the note to sufficient quality, it's not economical compared to a smaller orchestra, because experienced musicians do the same in real time, with 1 practice and 2 takes on average. The performing group gets it done in about an hour, while diddling a synth rendering can take weeks. Even though it's one dude or so, it's a LOT of one dude.
Plus, you risk "ear burn-out" from so many replays such that you cannot recognize quality anymore. One has to switch between projects and styles to keep their ears fresh, delaying the finished product.
Maybe the editing software eventually will get better and the computer can assist with more natural "guesses" to get closer to expectations to reduce customization, but at this stage if you want quality performances, synthesis is not fully competitive.