Well, the fundamental stuff is to have the fundamentals down. A lot of games do this okay, but it's surprising how many still seem to have issues with their controls, their UI, that sort of thing. That's tolerable for a game like Dwarf Fortress, it's just one guy after all, but a bigger production should never have these sorts of problems at this stage in the medium's lifetime.

Beyond that, the major thing on my mind right now is having a solid and fair sense of progression, one which rewards the player. When you make a choice, it should be between viable options most of the time - the difference between Civ 4 and Civ 5 is very telling here. In the former, you're making choices between what to build and where to spend resources, but pretty much everything you can get is useful in some way or another; the costs are generally opportunity costs rather than 'actual' ones (i.e. in the time it took to build that library, I could have had four swordsmen)

Conversely, in Civ 5, it's a game about choosing the least of evils. Building units costs you. Building buildings costs you. Building cities costs you. Building tile improvements costs you. etc. etc. You don't choose what is best for you, you choose what is least bad - and a 4x, or any game about building a great and mighty empire, should never punish you for that sort of thing.