The weakness Matthew pointed out is common to most (maybe all) strategy games: Before actually winning the game, there is a phase in which you surpassed the AI opponents in most or all areas, so your victory is basically assured and the act of achieving it is just kind of a formality. Since this is a weakness of the genre in general rather than GalCiv in particular, it´s nothing that "needs" to be fixed, but it would, of course, improve upon the gaming experience.
At first glanze, I see two ways to deal with it:
- Behind-the-scenes adjustments: Modern racing games often employ a kind of rubber-band system, that slows down AI opponents which are far ahead, and speeds up those who are far behind. Basically, the same could be done in GalCiv: opponents which are far superior to the player receive a malus to their economy, research, production etc. to allow the player to catch up, while those who are much weaker than the player receive a bonus. This way, the balance of power among player and AI races could be improved, and the game would be challenging throughout the entire match.
- Visible adjustments: Aside from giving the AI opponent hidden bonuses, the AI opponents themselves could be improved by implementing scripts that allow them to recognize the players superiority and find ways to undermine it. There are already some minor scripts of that kind in GalCivII, for example the AI refuses to trade weapon techs to players that have a strong militia. More radical scripts could, for example, have the AI form big alliances of 3 or more races all allied to each other to defend against the threat presented by the players superiority. Or they could have weak races who don´t stand a chance on their own surrender to stronger allies, to create a force that can oppose the powerful player. There are a lot of minor and major scripts that could be used here.
The advantage of the first approach is, it should be rather easy to implement and I think it would be quite effective. The AI would be kind of cheating, though. The advantage of the second approach is, that it´s visible to the player, making the game more interesting, and giving the late game a unique feel. It´s more difficult and time-consuming to implement, though, and might not be as effective.
One final note: something you must never do is "hurt" the player in response to his success (like in Matthews band of raging pirates idea). In a game, the player must always be rewarded for doing something well, not punished. Otherwise you risk hurting the players enjoyable gaming experience and motivation. Of course, if the player won´t realize the connection between his success and the bad things that happen to him it´s fine. So things like the pirates idea can still work, but you have to use them carefully and economically.
Anyway, the focus must always be on strengthening the AI opponents, not on weakening the player.