First, I cannot agree that realism / suspension of disbelief is the least important level of discussion, third after programming problems.
Quite the contrary, I think that game balance and suspension of disbelief are the most important, the most fundamental aspects of the game, while realism is just subsidiary instrument (one of them) of disbelief suspension. So there are 2 levels, and balance / suspension of disbelief are just 2 parallel parts of 1st level.
And, more of that, Aurora is a great and unique story telling SciFi game, not a game of great challenges (I have won at 1000% of difficulty with the same easiness, as at 100% or 200%). So, suspension of disbelief is much more important for me in Aurora.
Second, if we look at programming level, then we see that balancing of the multiple methods is a great problem itself just at that level. Current version of AI is already broken, AI just cannot use those multiple methods, and multipleyer is not an option for all players except, maybe, very rare maniacs. It's very hard to code good balanced AI. On the contrary, that's relatively easy to code interconsistent model, not a patchwork quilt of different methods and objects. And that interconsistency is a key instrument of disbelief suspension.
I think, that for the first versions of C# Aurora it will be good to have some simple mechanics (to have working AI at all), even if it will be some "one button strategy". We'll get our pleasure for long months just finding that strategy, while Steve always will have an option to add some new feature.