Question How mlagents tackle the curse of dimensionality problem in multiagent reinforcement learning

In a very popular article called "Multi-Agent Reinforcement Learning: A Selective Overview of Theories and Algorithms" it has been said that "the joint action space that increases exponentially with the
number of agents may cause scalability issues, known as the combinatorial nature of multiple agent reinforcement learning (MARL)"

In another article as "A survey and critique of multiagent deep reinforcement learning" says that "One way to tackle the curse of dimensionality challenge within multiagent scenarios is the use of search parallelization". Do we have anything like this in mlagents ? Does it count as parallelization when agents work in the same scene but in different environments ?

Thanks