I plan on it yes.
And no, small hits here and there won't damage the play speed. Hell, if I can use a 1024x2048 cycling animated texture for the water without anything but a load-time hiccup, then I can still do just about anything else I want with basic decals. And if mine can handle it, I am certain everybody else can handle it. That is one of the beauties of building everything with an ancient device '>
Vanilla used a lot of transparent textures. The key problem with transparency is when you cover a large portion of the screen with it. What I plan on doing is covering a larger portion of the base texture with something else, like vegetation. And keep in mind, you won't be able to go examine large portions of the rocky masses, so you won't ever get to see the texture lines in game. It won't be naked as shown in the images so far, but fully dressed with a high polygon count. The trick is making low poly shadows for everything, and breaking the key features into facing groups which don't interfere with each others shadows.
Take for instance the line between mountain bases and ground bases. Right now, the entire section is a single model, with a horrible shadow issue. It lights wonderfully, almost perfectly. But it's only one piece, and that causes a lot of issues later. What what you do is get a non-transparent texture to straddle that line, which creates two more lines. Then you ride those lines with a very thin strip of transparent decal specific to the transition, reducing the overall calculation of transparency. And then, you cover anything unsightly with a sub feature, like a plant, tree foliage, rock pile, etc.
Same goes for water/beach transitions. Three basic textures, one or two transition strips, and then details.
Keeping those transparent transition strips small is key. Even though a large portion of the image may not be transparent, the larger the image is, the more calculation will be done. I shouldn't say the larger it is, I should say the larger the polygon using it. That is more accurate, because it isn't how much you are mixing, it is how much you are mixing it with, and at what levels/distances. If you constantly have to blend large regions in front of the camera, from all the way back, all the way up to the camera, you are doing a lot of calculation in a single screen draw cycle.
What your output will actually be when this is all finished, is a completely seamless landscape, very much like that you can create with NWN2, just without the bump mapping....well for me anyway, since I can't do bump mapping on this computer. Bah.