Super cool stuff!
Did you use the same sweet spot theory in coding your bot for engagements later in the game? I can definitely see how this sweet spot logic could help you set up a sort of wall that would just smother advances from a more naive opponent.
If you made a map of regions approachable by one ship, assigned the enemy ships' ids to regions they can approach, and then solved an assignment problem to maximize the number of enemy ships you might gang up on, you could set these situations up all game long.
Would that lead to too much stalling behavior? If you did this, would you want to prioritize regions of the sweet spot near the opposing ship, to quickly close in on and eliminate enemy rabbits/solo-invaders? Getting close to the ship seems like it would increase the odds that it would navigate into combat range. Spreading ships out inside the sweet-spot might help your ships cut off escape routes for runners.
I started coding my bot with an eye toward creating coordinated 4-5 ship wings, and it seems like this would have been a great continuation of that idea. Unfortunately, I sort of lost steam a month ago, and never really got wing behavior working right.