I’m looking for guidance on the right approaches to achieving this task: We want to use a load balancer that handles initial requests from users, selects the most appropriate server based on user requirements, and performs servers health checks.

After the initial allocation by the load-balancer, subsequent communications should done directly between the user and the assigned server, bypassing the load balancer.

Is it possible to achieve such a task using NGINX or HAProxy to enable this direct communication after the initial request from the user?
How do you maintain session persistence and security when bypassing the load balancer in subsequent requests?

Thanks in advance.