Deep strategy. The first step was separating attack and damage, dodge and defense. The second step was making weapon and armor choice meaningful beyond "more is better." I quote: "Now, army composition is a major consideration, and the composition of opposing armies is a major factor in winning tactical battles."
I think this is a major point, but I think you are a bit to vague. Right now, Mixed armies of normal units is not exactly meaningful, and once combined arms becomes more meaningful so will tactical combat gain a lot of ground.
Since v1.1, it has evolved by removing functional spellbooks (which just felt tacked on, let's be honest here) and by making the schools contribute to high-level stragtegy in meaningfully different ways.
1. Also, the removal of the "preselected shard type" system in favor of a "generic shard which upgrade to specific shrine type" system. What can I say, the reviewers absolutely loved this one.
2. Serious scalability of magic through stats increases.
I disagree that functional spell books are necessarily a bad thing. Honestly, I think that normal units need to become a bit more "magical" and the spells simply needs to function in balance with this new magical existence. As for your generic shard idea, I think that generic shards could be an interesting addition, but I think that it would take away more than it added. Right now, getting shards of a type you want is challenging and requires a player to engage the world in a meaningful way. Personally, I think that shard power should be able to be traded diplomatically, so a person with 4 air shards could trade one of his allies for a fire or a earth shard to fit his strategic needs.
The hiring of a statistician to consult on the math behind the game mechanics. As soon as Brad Wardell realized that the game mechanics could be modeled graphically, the balance of the game improved dramatically. When the numbers work before the coding is done, it freed up the programmers' time to add content, saved frustration all around, and improved the balance of the final product.
This is just unnecessary. The best 4x games are not about number crunching balance, but about the feeling of being part of a rich and vibrant world. The only thing this would directly accomplish is making a game which can be easily minmaxed.
The inclusion of more base spell effects. There were only so many ways to write a spell with less than 5 effects, after all.
I agree with this, but I think that new building blocks should be well considered before being added to the current set.
# Unique tech trees for each race.
Unique personalities for each race.
This would definitely be nice, but I don't think it is necessary for the game to be more fun. Something as simple as a couple unique units and buildings could do wonders for the game without the hassle of making completely unique tech trees and personalities for each and every race.
Separate damage types and resistances.
I actually disagree with this. This was something that was done in GalCiv 2 that I actually found to be less than stellar. Honestly, damage is damage no matter what form it takes. What makes combat interesting is how you deal damage and what effect it has on the unit. For example, A maces and hammers could damage armor as well as damaging a unit, or a shield could be ineffective at blocking attacks from the rear. This would add far more strategy than simply a number crunching system based upon various damages and resistances.
One of the greatest boon to giving the game more interesting choices would be a passive enchant system, allowing you to assign enchants to weapons and/or armor. These enchants would then passively affect the unit, for example, giving a chance when attacking to do extra [fire/ice/chaos/true/*] damage. Ideally, these would receive minor bonuses from shards as well (higher chance to go off).
This is a fairly good idea, though I would expand this to say that spells should be expanded to have a sub target. So then, we can target weapons, armor, items, or the unit. On top of this it would be nice to have a party target as well.