Design real-time modules communication and overload handling

You are designing a real-time system running on a single machine with two modules (e.g., producer -> consumer) that must communicate with low latency. 1) What communication mechanisms would you consider (in-process calls, threads + queues, IPC such as pipes/Unix domain sockets/shared memory, gRPC/HTTP on localhost), and how would you choose? 2) If one module becomes saturated (cannot keep up), how do you prevent the system from collapsing? Describe strategies such as buffering, backpressure, load shedding (dropping packets/messages), prioritization, and what metrics/signals you would use to trigger them.