You mostly do not need to worry about pitfalls of concurrent processing (races, deadlocks, livelocks, resource sharing, etc.). The fact of cooperative processing takes care of such issues by design.

Authored by embedded ML specialists with extensive experience in ESP32 voice recognition architecture, TinyML optimisation, ...