Investigating Client Timeouts

Our clients are seeing sporadic bouts of HTTP timeouts when trying to connect to our service. The timeouts become more numerous with increased overall load.

The service is not overloaded, at least not according to the metrics (cpu, memory, threads/connections available, etc) I have at my disposal. I suspect this is related to our load balancer configuration (we have cross-zone routing and client IP preservation both enabled; AWS docs say this is poor form and leads to client connection issues) and I would like to know if there is a convenient way to access logs pertaining to the connections handled by the load balancers. With these logs, I hope I might see that connections are failing for one reason or another. Perhaps theres a better strategy that can be recommended?

Edit: Access logging is not available due to our listener being a TCP not TLS listener.