I also agreed with this post like 98%. They have identified all the same, most important issues with the game that I have (except I'd ignore the lighting one, the engine is ancient by this point, it looks good enough, and I'd much rather processing goes to better/more simulation of gameplay mechanics than graphics) and almost all of the proposed solutions would solve these issues given they could be feasibly balanced without breaking ****.
In terms of prioritisation, the adventurer>merc>vassal>king route is the only way to play the game and get decent hours of enjoyment so making that complete should be the first port of call. Early/mid game is pretty good for now, but later on, as a king with a decently sized kingdom, it is a slog. ESPECIALLY THE CONSPIRACY QUESTS PLEASE LISTEN TO THIS GUY - Increase the amount of impact that the conspiracy quest completion/failure has on the conspiracy meter, decrease the frequency of conspiracy quests dramatically, 1/4th the current frequency is where I intuitively feel it should be when you're in constant late game wars, and increase the size of the forces you have to defeat when you decide to do the quests at minimum. But that is only for story mode which people aren't playing anymore until it gets fleshed out and said story has meaningful impacts beyond the near identical sandbox experience of building a kingdom.
As far as sandbox goes, the above "Wars and Armies" and "Diplomacy and Politics" feedback that OP posted should be treated as TW's in-house bible until something like them is integrated.
After that, TW should vary the NPC Lords' decisions using the established trait system, some of my following propositions could be light on balancing given the following preconditions:
As long as each trait value is redistributed among the lords of Calradia with an even, equal chance within each faction, and the effects of deviating either direction from the equilibrium (0 trait value) for each trait are inversely proportional to the opposing side of the equilibrium, thus making the statistically average lord's decision stay identical to the status quo, and given that the equilibrium for these values IS actually 0 (all traits atm are currently either from +1 to -2, or -1 to +2 so they would need to be changed to have an equal variance both positive and negative, i.e -2 to +2 or -1 to +1), and each trait score of 0 is set to be identical to existing vanilla lord behaviours, i.e. no change.
An example of a "light on balancing" change using this principle would be if the generosity trait could influence stinginess in barters by a percentage increase/decrease, including the cost of recruiting a clan to your faction. This is a "light on balancing" option as the negative would be directly inverse to the positive in every way, so you only have to balance for an acceptable amplitude of deviation. Say the vanilla average cost of recruiting a clan is 1 million denars, the trait system could either increase or decrease that cost by 15%. Assuming they are an equal number of generous and stingy lords, a player blindly buying whoever they can, acting how one does in 1.7.0, will pay the same amount of money through their playthrough, those who read the game mechanics info pages in the encyclopaedia will see that approx 1/3 of clans are gonna be cheaper to recruit, and their rise to influence can in advance be spent cosying up to a lord that doesn't want large sums of gold.
Considering that the relation system exists though, it would make sense that a generous lord who is also good friends with their liege would ask for MORE coin to join you, and a closefisted lord would be more "generous" to you, ironically asking for LESS money, IF they despise their liege - the implication being that they'd seek monetary gain from means (i.e. you) other than their current liege. So in the case of clan leader recruitment, an additional inverted relationship equation could be added for better in world context.
The honour variable could directly cause troop composition preferences, low honour decreases cavalry recruitment and increases archery recruitment, 0 is vanilla (which is normally just infantry dominant), and high honour is more cav, and fewer archers. This would create a tangible, and important differentiation between lords, yet each faction's total military force would be statistically balanced/holistically identical for overall troop type distribution.
I suppose there are actually multiple ways that balance could still be interfered with if either of those +/- ends are disproportionately stronger than each other when comparing individual parties, but there are multiple ways to tweak this so that whatever balance TW is going for could be maintained.
Percentages for these troop preferences could be adjusted, asymmetrically if needed (maybe archers are stronger than cav troop for troop so archer preference could increase by 20% while the opposing trait value's cav preference goes up to 30% extra) and if the average lord's strength is raised or lowered in the same direction by the +/- honour values, the preference for non-infantry troops at 0 honour could increase/decrease OR the average infantry total at the +/- ends could decrease as a counterweight to the specialisation preference.
I'm not sure how cav archers would be defined in terms of honour, but horse archers seem pretty dishonourable to me so I reckon they should count as archers for this equation.
When an ai army/lord is attacked by the player's side, a low 'calculating' trait score could give like a 20% chance of the enemy ignoring a defensive position and pushing right up to your forces to charge you because they've 'miscalculated'. There really is no easily determinable, equally opposing force for this equation to be "light on balancing" so the holistically identical approach doesn't work here, but at this point of the mount and blade series' long history of predictably identical tactics, I think ANY variance in ai battle behaviour would be a nice treat. So maybe a high calculating trait score also has a 20% chance to push up (so they're not next to the edge of the map), but for the purpose of actively trying to use as much of its ranged ammo at 80% of that ammo types maximum distance (Shooting from xbow distance, then bow distance, then closing in to like ~20m for throwing weapons, and when the ammo runs out, the enemy lord executes the retreat command (hence why they should push up first - to give you time to chase them down). The gameplay loop here would be trying to crush a force that is going to slowly move away from you while firing, eventually running away if you're too slow to entrap them with a killing blow. If the enemy lord escapes with their troops using this tactic, they will continue to use it repeatedly for all ensuing battles in the next 12 hours against the player, to prevent the player simply identifying that this strategy is being used and leaving the battle to start it again without the strategy engaged (as it's only a 20% chance of occurrence, in this example).
This means that outnumbered forces with a calculating general will use the same "shoot as much ammo as possible then press tab to retreat and restart the assault" tactic that the player can (and does) exploit frequently (A good additional change to that exploit btw would be to make the player have to retreat every troop to the border of the map as well - and only being possible after a couple minutes to allow the enemy a counter attack).
Alternatively to my suggestion, TW have backups of obsolete and inferior ai tactics from earlier game versions, and possibly some versions that are even more competent than in 1.7.0 so they might already have options to straight up implement dumb/smart ai tactics that could be tied to this calculating trait.
The Valour trait variable could sway ai lord's confidence in the balance of power, daring generals being more willing to engage in potentially tenuous engagements and cowardly generals only wanting to engage when the balance of power is more in their favour. You would only need to discover a suitable amplitude of deviation to balance this.
The mercy trait of a lord or even army leader could slightly sway strategic world map decisions, i.e. more tendencies for raiding villages, villagers, and caravans with a low mercy vs field engagements with the enemy's forces, besieging and (because the previous two are more productive than raiding) maybe releasing the occasional prisoner for mere relation gain (like the player can do after capturing enemies).
These changes would mean that enemy lords are not all equal, identical foes and the player could theoretically identify some of the traits of a lord without even checking the encyclopaedia, but just by fighting them a few times. It also gives long term strategic thinking to the player. Which personality types should I train up to grant land to? Who should I try to gain relations with? Which enemy lord poses a bigger threat to me and for what reasons? Their tactics or their army compositions? Their high cost of conversion? Should I cripple this honourable and generous lord's villages to stop them bringing cavalry against my archer focused army or take away their lands so they dislike their liege and drastically reduce the price of conversion?
After giving the lords some personality, I'd focus down the lack of RP options in your RPG game. Being able to conquer the world economically through mercantilism would be a major major selling point when in addition to military conquest, but seems difficult to suss out, the economy of Bannerlord seems to be the most fickle and unwieldly beast to manage.
Banditry options based on inciting unrest and riots so that you can take a faction's land without declaring war would be perfect and a basic version of that would be easy to prototype cause all the ingredients are already there. You have a conspiracy quest points system where roguery missions are presented to you at regular intervals. When enough points are gathered, you incite a rebellion in a castle or city, which you can already freely take without declaring war on a massive faction. Adding this would be an net improvement and what would you need? Three equivalent quests to the conspiracy quests, but focused on doing naughty things, then just follow the template you already have for the main story, but instead of triggering factions to declare war on you, a rebellion is triggered in a relevant area (maybe within your starting cultural boundaries?). Working it in as a functional narrative would take time but the prototype could be bing, bang, then boom!
To the extent that the early/mid game falls flat I think is mostly down to heaps of quests only being worth doing for their first time novelty. You get unavoidably punished to some degree for doing most of the quests - pretty much all of the ones that aren't "kill thing here", which you're doing anyway. This is very anti player, anti video game heuristics, and anti sense. When I accept a quest, then actually follow through with it, I can get an extra recruitment slot with the person I did the quest for, yet potentially lose two slots with the person (or sometimes, god forbid, multiple notables) I've just insulted. A small gold profit never makes up for losing recruitment power, ever. Stop thinking it does.
May be some errors in this, I'm very tired and will proofread later.