In the eHPC (embedded high performance computing) market, developers are focused on domain-specific application development (e.g. radar processing, signal intelligence, autonomous driving). These domain-specific problems require substantial algorithm expertise (math, physics, etc.) not related to communication.
The resulting applications may be compute-intensive such that in order to achieve real-time, the algorithm must be distributed across multiple compute elements (e.g. CPUs, GPUs, FPGAs, ASICs) that make up a heterogeneous system. They must also integrate high throughput I/O devices like video, lidar, and radar sensors as well as high speed storage. Distributing the processing and integrating I/O currently requires multiple nontrivial communication interfaces (e.g. sockets, verbs, shared memory, semaphores, conditional variables).
Overall, eHPC developers are immersed in the algorithm development (the primary focus), but when it comes time to distributing the application, point-to-point communication is typically a secondary focus.