Spore had great *visual* unit customization. What annoyed me to no end was that the vast majority of unit/building/ship design decisions had no impact whatsoever on gameplay, aside from a few extremely limited variables. I couldn't put a weapon at the end of a very long creature limb for added range, or position vital parts in more defensible parts of the body. A creature I made with 8 legs would move around in exactly the same way (in terms of mobility) as a creature with one leg. It's a prime example of a game with potential for incredible depth that failed to fulfill most of its potential due to what seemed to be to be a fear of not being able to reach the absolute broadest audience possible. (Mind you, it was still a good game, and for casual players maybe even a great game, but I felt that it really missed an opportunity to be great for everyone)
I much prefer the "Onion" approach to complexity. Put the complexity in, but make much of it automanage itself unless the player "opts in" and make sure that the game is fully playable and fun without diving into the automated complexity. It's harder design since you have to design complex systems right alongside of more easily understandable abstractions, but it scales so much better. It also lets a player get involved in the game without being overwhelmed and then gradually discover the richness of how they can interact with the world.
It's much the way that we learn things in the real world. We talk about concepts like gravity on an abstract level by talking about Newton and an Apple, and things falling - in terms that children and people with no scientific background at all can understand. Then when someone is ready they learn the actual physics of it - but that initial stab at the physics often doesn't include any calculus (which is abstracted away). Then maybe that person learns the calculus... and so on.
The end result is that if you asked 10 people with varying scientific backgrounds and affinities for science what would happen if you knocked an apple off of the table, they could all give you a useful answer - but each of them would be able to approach the problem in a way that fit their individual knowledge set.
That's just a longwinded way of saying that I really hope potentially deep and rich systems such as unit customization are done in a way that fully exploits that potential, without fear of alienating potential players - but that they should should also be built with completely transparent abstractions (defaults, automatic configurations...) that allow a player to bypass all of that complexity if they so choose.
If I had my druthers as to exactly how complex a unit customization system would be... Well, probably very complex. There would be city-by-city variations in effectiveness and specialization, weapon skill ratings on a per-unit or per-figure basis for different types of weapons, the ability to equip and train units for different types of combat, a robust unit experience system with a large number of potential effects - and that wouldn't even touch on the large variety of differentiation that might be attainable from equipment quality. Do I expect to see anything like this? Of course not - it might be just a bit too much under the hood for even a really, really good abstraction. However, I'd hope that there would be hooks for modding this sort of thing in at a later point. From what I've seen, I think Stardock ascribes much more to the masked complexity approach than the "dumb it down" approach that Spore took, and so I have decently high hopes (and not merely for unit customization)
tldr: Don't design complex systems in fear. It's possible to abstract a complex system in a way that makes it easy to use for beginners. Spore is an example of what you get when you design with fear (Fear and a very high budget, though.)
