But that could be limited through proper AI:
If the AI properly garrisoned cities they would be more difficult to take.
Another way to do it would be to improve the way treaties are negotiated. For example when setting the treaty:
- a cap could be set on the number of units traveling through the neutral territory
-a specific path through the territory could be set, if the player deviates they receive a warning then they are attacked
-a player army could be shadowed by a AI stack as it moves through neutral territory, if the player attacks the AI cities he/she would have to fight the garrison and the escourting stack at the same time
I could probably come up with a few more ideas, but the basic idea would be to have the AI react as any reasonable player would. If a neutral stack enters our territory we would either attack it or if it does not present an immediate threat then we would monitor it, I see no reason why an "evolving" AI could not mimic the behaviour.
regards,
ct