ISO 2D engines and the issues therein are fairly well understood, however doing ISO in 2D can be more complicated in some ways, simply because you have to deal with depth sorting and whatnot -- This isn't especially true if you only have 1 ground and one object layer though -- the trouble comes when you want to be able to stack layers of ground in order to provide a sense of height, and particularly when you combine that with "tall" objects attached to the base layer (eg. tall trees). It's not mind-bending, but you have to be aware of your drawing order, the size and shape of your objects, and all the weird corner-cases that can crop up.

In 3D, you can usually just use a depth buffer and it will all be sorted out for you. However, you can't simply do billboards for everything in 3D, at least, not without baking some kind of pre-computed depth offset into the texture and compensating in a shader (which might prevent some optimization). If you do ISO in 3D, its probably better to map a texture onto a simple 3D shape that approximates what you want -- some things certainly can be billboarded, but larger, solid structures like buildings probably should not be.