Reducing the fidelity of the simulation (at a distance) will change the outcome of the simulation.I'm no computer scientist but if each layer only has a few hunded objects and they don't have to be constantly updated every second it would be feasable to simulate their interactions.
The region where the player resides would be simulated differently than a far away region. The ai would come to different solutions of what to do when trading.
Thats a very general problem with the approach in LT (having no "limit" in scale) , and will either limit the total number of fully simulated systems/ai traders, or make the simulation diverge.
So either the trading decisions have really low updates, so a normal PC can keep up with all the ai decisions, or there might have to be pauses to batch-process far away systems.
"please wait while updating the world"
I fear, the easy way out will then be to simply emulate the outcomes of complex ai decisions, and script market cycles to be more of an entertaining gameplay (than a full simulation).