I have hardly played the mod, and then only the earlier and limited Workshop version, but I have been following the thread and my experience is consistent with what people are reporting.
I agree that AIs being so aggressive against the CS is a fun-killing problem.
The AI doesn't value luxuries in its conquest of city states so it wouldn't change the frequency of killing... My main purpose with this would be to also help the player in having more randomness in CS luxuries rather than have 5 allies all with the same lux. In a way an attempt to raise the value of the alliance.
I don't see that raising the value of CS to player helps the problem with the CS being killed off. CS allies are already quite valuable. The player already has sufficient motivation to Liberate CS whenever possible. Buffing the CS ally benefits only makes it more frustrating that there is nothing to do for the CS that are not neighbors!
You have explained that making the AI think is not really an option. Can the AIs be made to value pledge-to-protect more highly? I would like to see the AIs fighting against each other more over the CS -- and see the AIs liberating CS more. I would like to see the AI hesitate to DOW a CS with a pledge to protect because they do not want to war with the allied AI. Is that sort of thing reasonably easy to implement?
The root of the dilemma, as I understand it, is that the same algorithms that make the AIs more aggressive with the player cause them to be more aggressive with the CS. Is that correct? The CS are a soft target, and the AIs like soft targets.
I've got a better idea about CS's and protection... If Mongolia keeps conquering CS's, shouldn't I gain influence with a CS for killing Mongolian units in and around that CS's borders?
That is a very compelling change, but more for the main game. I have no idea if it would be easy to implement and it only really helps with close-by CS. Neighbor CS already provide the motivation to liberate. Now, liberating is not as good as helping them not die in the first place -- but the Liberate-the-turn-after-first-being-conquered exploit would still be stronger than helping the CS not die when they first get DOW'd. So I do not your interesting idea helping this problem, since I think the bigger issue is with CS halfway across the map.