There is no doubt that Conan servers could run better, and I am sure some tasks could be offloaded to other threads, like NPC AI is usually a good candidate. But it is really hard to estimate how much that would help, without rewriting the whole game. Evidently a lot of bad decisions were made along the way. I mean, we already see the NPC AI freezing when there is enough players online, so I imagine they made it very low priority… offloading it would not make the game feel much less laggy perhaps, but it would ensure that the NPCs function, which is… something.
My understanding is that big multiplayer games using the Unreal Engine (and similar) basically need to use fully custom code for their server executables, if they want good performance, because the basic server code it ships with is typically only suitable for a small scale pvp shooter (such as the cancelled Unreal Tournament title, if speaking of UE4). If your game differs much from that template, you’re going to have a bad time, and that’s exactly the situation with Conan since Funcom didn’t have the knowledge/experience for such things. Swapping to UE5 would help Conan servers very little, if at all (but I am sure it offers better client side visuals though).
I do not think any game in existence splits the handling of players between threads the way you describe it, for one single map/instance. The reason is that it’s not as trivial as it initially sounds - one thread might finish things before the other (and thus get stalled), or the calculations produce a different outcome (just think about the RNG elements in many games), and suddenly you’re totally screwed when anything happens around the edges of the map areas, where players who are handled by two different threads meet up. Which thread is right, which thread is wrong? Or are they both right? Or you have a master thread who calls the shots, but then if a lot of activity happens with players from different shards… you suddenly find yourself with a new single thread bottleneck.