I think a lot of these ideas could work to improve the AI surrender/death issues. Whether a surrender routine is initiated by the AI, or human player, the AI should continue to fight and play normally until both the AI and the human player agree on the AI surrendering. And while I think this is true when there is only one (or only one remaining) AI opponent, it is especially true when there are multiple AI opponents in a single player game.
The complexity of a FFA with multiple AI can decrease quickly if AI that are in control of multiple worlds and still have a fleet get discouraged and quit early in the game. I understand that in some sense their position probably had become hopeless, but really, who cares? Having to decide whether to allocate forces for wiping them out or to concentrate on the stronger AI enemies is the kind of tactical decision that makes the game fun. And if the weak AI is lucky and the focus of fighting turns away from them, or if I, for instance, enter into a cease fire with them to have a quiet flank, then perhaps they may have time to re-group and put up a reasonable fight.
On a minor related point, do weak AIs enter into cease fires and peace treaties more readily? You would think that would be be a good idea, since it would make them more viable, and reflect a realistic kind of diplomatic decision-making, as well.