Hi @CarlosJong,
generally, nothing speaks against using Node-RED for that purpose. But it depends on the specific use-case, of course.
At work, we are using it in production systems (also not IoT, but complex production lines) as one part of our software stack. Mainly for the communication between PLCs and our backend services (those are not driven by Node-RED).
It has been used in several customer projects now, and - compared to the previous solution - we have gained a lot of flexibility.
Application performance is expected to comparable to any other Node-JS application, which runs a single-threaded main event loop.
I'm not sure why you'd want to make changes to the Node-RED core for your use case? 
But regarding documentation and pointers, I think the core developers could help you with that.
