MultiAgentic RAG swarm related discussion

Hi guys, I have one question related to MultiAgentic RAG and swarm use cases. Let’s say, if you have a Main agent and under you have multiple agent and more sub sub swarm agents. Assume that that might create 10-20 agents hypothetically. If you create this amount of agent, swarm sub agents is it scalable in terms of production level code? Or not? Also, how can one make sure agent have good memory, better performance, optimize etc? Any suggestions or thought to discuss?