Here is my two cents about some additional feature ideas, that i think would help prevent snowballing naturally, rather than changing values to put balance between factions forcefully. These feature ideas aren't actually new ideas, some are bulletproof features that have been implemented in more than dozen games, where army composition and market based economy exists.
Demographics penalty for rebellion: There is already rebellion in the game with the last patch, but i guess it's implementation, naturally, is lacking. This is not unexpected, it would take many patches to perfect it. Demographics penalty for rebellion would be that, if, for example Battania has 20 Battanian fiefs and 15 Vlandian fiefs, chance for one rebellion in Vlandian fiefs would be higher than another scenario with 20 Battanian + 5 Vlandian fiefs under Battanian control. There could be some math to calculate additional rebellion chance for those Vlandian fiefs under Battanian control. More different cultured fiefs one faction gets, fiefs with that culture in that faction would get additional rebellion risk point. A lot of games uses this type of features.
How would that prevent snowballing naturally? If one faction expands aggressively, they would face higher rebellion risks than others who don't expand as much aggressive. (Assimilation feature would be needed to balance this feature, or no faction would get so big even after centuries).
Faction aggressiveness penalty for rebellion: There is already aggressiveness as a kind of variable if i remember correct. If one faction is warmongering so much, that faction's cities would have higher loyalty loss penalty during and after the war (effect would gradually decrease when the war is over). This would put additional stress on lords to side with peace. Some games use such features for rebellion calculations.
War targets: A lot of games have that, factions would have their own justifications and targets to achieve, before declaring war. How would that help prevent snowballing? There are wars that start in the beginning of the game, and continue like they will never end until one is no more. I guess war targets would add a lot of color into the game, and help prevent snowballing naturally.
City States: Rebellion feature was very needed, but i think it could be more fun to see city states, like one city-factions. If a rebelled clan fails to join any faction, they would create their own faction. How would this prevent snowballing? It would add more balls into the diplomacy pool, imo this ca naturally slows down the overall expansion of factions.
Pike bracing units (Pikemen): As a 5th unit after inf, archers, cav, horse archers, they would have their own division. I remember reading that TW already plans to add Pike bracing units. So, instead of changing cav ratios of factions to put balance, AI battle calculation would count Pikemen as a negative variable for enemy cavs (and non-Khuzait factions would have more Pikemen naturally). How would that help prevent snowballing? Imo, it at least would help prevent certain snowballings where Khuzait is the one leading the race, and there wouldn't need to play with the cav ratios in armies. I personally find Khuzait realistic for them to have many cavs, even tho i don't like playing with them. Pikemen unit would be a natural antidose againt cav-based factions. (also shock troops need to have their own divission too, instead of getting stuck with shield infantry).