Yeah. The built-in Ignition redundancy feature is NOT what you want for load balancing Perspective. I went down that path briefly and tried to explain some of my pain over in this thread:
Your multiple FE servers should have nearly identical configs, not be aware of each other, but all talk to the backend Ignition servers via the gateway network.
You will want the AWS load balancer to be sticky in some form. You can use a simple one based on just source IP, but you can get fancier and look for the JSESSIONID cookie if you want.