There are core issues and then issues that would just be trade offs if the core issues were solved.
#1 is that having high 3 to low 6 digit numbers of AIs means that with real time gameplay you just can't have the AIs make complex, much less complex *and* good, decisions. You just can't optimize enough. Paradox is a massive "AA" company and they can't manage to make good idea even with how streamlined their games are mechanically. Now imagine if the AI needed to have character and long term planning and such. Blech. You also have much more trouble multi-threading in real time in a strategy game.
#2 is that you can't have complex "events" with multiple participants. If you read the substack blog posts about social occasions in Axioms you can instantly see that that would be totally impossible in a Paradox game. The same for War Councils which are basically themed social events for a specific purpose. This is one reason the "court" is so unconnected with the rest of CK3 for instance. And yet the events are still super simplistic and unconnected to the outside world.
#3 is that Paradox games are incredibly timer based. You have a lot of "dead" time. Turn based games can have dead time(turns) as well, if you don't balance the action economy, but it is a lot easier to fix if you really want to.
Most Paradox games, especially EU4 and maybe Vicky3, are functionally idle games, except they have a map. If you play something like Arcanum/Theory Of Magic, you'll easily analogize "furniture", "rooms", and "upgrades" for your "tower" to "ideas", "decisions", and "estates" in EU4. EU4 famously has "mana" like a wizard tower idle game, also.
I got off the Paradox bandwagon 2 generations of games ago ago (EU2/Victoria 1/HOI2), so I'm not sure what the new games play like, I was more thinking of the military AI in especially the first Imperialism. It was pretty amazing for it's time, especially the way it was basically impossible to sneak-attack. The AI would never get caught out of position like even the old Paradox military AI (which had an easier time because of smaller map sizes and complexity) would.
When you started your stack marching towards a province in EU2 or whatever with fog of war turned off, you could literally see the AI skip a few ticks before they realized you're on the way. With those time lags for each single decision you take adding up, it's no wonder players can easily run rings around the AI.
With Imperialism for example, if you were playing Germany in the 1881 scenario and wanted to sneak attack France, you move your troops to the border and declare war, which would start the next turn. But the AI sees you moved your troops to the border so it mirrors your deployment meaning you can't catch them out like in Paradox games. There's no military access mechanic, so you need an ally or multiple border provinces to force them to divide their forces, there's a battle stack limit so you can't just hope to throw everything at them in one go, or you get naval superiority and establish landing zones to make them protect their coastal provinces.
Which is another thing it did really well compared to old Paradox. If there's no naval invasion physically possible that, they won't waste troops to garrison coastal provinces, if certain ones are eligible to be naval invaded that turn, then they will send troops there then and you can't sneak a landing in because there's a turn's delay between the ships arriving and you being able to land troops, but they also reprioritize their navy to break your landings, etc.
It was overall very reactive, in the way that AGEOD games also do a great job with the campaign AI. The gimmick in Imperialism though was that (for the player and the AI both), distance between connected land provinces didn't matter, so you could move troops in mainland France around anywhere else in mainland France in 1 turn, etc. AGEOD doesn't do that (and has individual calculations for how fast stacks/units can move) but it's still hard to catch their AI out of position.