Ok, I got you wrong then on your initial post, this I can understand, compromises are never really satisfying. But just one question there, shouldn't the design be mostly finished before you start developing? I wouldn't start a big web application before I have a good technical concept. Of course there is always something that won't work out and then you have to find a workaround or a different solution. But designing while the engine is developed sounds like a bad idea to me. Anyway, you have more experience in these things and I won't criticize the way Stardock develops games, it is just my point of view how I would approach it.
Before you write a single line of code the design should be done. Elemental was no exception.
What was the exception was that the platform we were developing on wasn't as far along as we had hoped. These issues, if you're not in software development, sound very serious but it's fairly common -- except that Kumquat was LESS far along than we realized (mistake #1).
To use a pretty famous example, in The Empire Strikes Back, Lucas bet the farm that the puppet (Yoda) would be believable. If Yoda didn't come across as believable, it could have sunk the whole movie. In Elemental, there were a number of design elements that just weren't doable. When the original concept on how tactical battles would be done proved to not be doable (continuous turns) the switch to the turn based system was something that seemed reasonable at the time but really ended up damaging the game. Similarly, the spells (which were originally very similar to what's in Master of Magic) proved unable to be supported by the engine (and are still not supportable in v1.09 but is something we've increased priority on -- and before someone lambasts about this, bear in mind that different people love MOM for different reasons and for me, the magical spell variations was not the priority because MOM's AI never really used them and AI trumped spell power Mistake #2). Instead, that time went into making a powerful quest system which, works but for reasons that deserve its own post mortem hasn't yet been properly made use of (mistake #3).
Now, one could argue (with the benefit of hindsight) that we should have just sat back and waited for the engine to develop until it could match our design. The problem was that we really had no idea how long that would take. So, being the engineer that I am, I modified the design to suit what the engine could realistically do and believed it to be "good enough" (mistake #4).
Now, bear in mind, and this is where people will disagree (and there's plenty of people who think GalCiv II wasn't a great game based on what I read so bear that in mind as I write this next part) but if Elemental's engine hadn't had technical issues, it would have gotten pretty reasonable reviews. We had multiple former major game magazine editors on staff playing the game and giving their feedback. And having made a lot of games over a long period of time, I am pretty convinced that had the game not had the technical issues we would have been pretty well off in terms of reviews. Tom Chick and Troy Goodfellow would still have given it negative reviews like they did Civ V for much the same reasons.
Galactic Civilizations II was a huge commercial success and a critical success and it had techs named "Lasers IV" and has an economic system that barely makes any sense even to me. But when it was reviewed (weeks after release) it was very solid.
You always make compromises in game design based on what the engine proves realistically able to do. This is true whether you're making Elemental or Civilization or Left 4 Dead. Stardock made one of the expansions for Starcraft back in the day and our design had to be altered based on what the engine could do. It's always the case.
For Elemental, the failing was a combination of the engine being technically problematic in the wild combined with fundamental misjudgments on how the design compromises would affect the overall game play.
So to work with our analogy, not only did the Yoda puppet not work, but it actually caused many in the audience to have seizures. 
I'm over simplifying this of course. There is no silver bullet issue here. There are so many different challenges and decisions that could have been made differently. Heck, me choosing to spend the summer working 100+ hours each week (according to our security system I averaged 103 hours each week from late May until August 10th) led to all kinds of catastrophic choices being made (mistake #5) that only seem obvious with the benefit of hind sight and rest. 