QNX has a much longer and deeper history
While most seem to date their experience with QNX to the cute 1 disk boot of the later Neutrino kernel, QNX actually goes as far back as to the 80's, where originally it was a fully deterministic realtime microkernel system written entirely from scratch by Dan Hilderbrant for x86 PC hardware, and hence also predates the widespread use of things like ethernet. It was often used for industrial control (including nuclear power plants) and factory automation at that time.
Originally the unique feature of QNX was it's ability to treat micro-kernel system calls either as "local" messages between processes on the same machine, or distributed over a (arcnet) lan, in an entirely seem-less manner. Hence, one did not need to know where the actually computing for a system call physically happened. In this incarnation it support x86 real mode and later 286 "protected mode", as well as a C library that was generally "compatible with" posix/usrgrp libc "visible" standards. It's other primary characteristic was very high performance, thereby demonstrating indeed what the later Mach and GNU Hurd failed to achieve with micro-kernel architecture was actually possible.
Later came QNX4, which offered a more fully posix compliant interface while retaining it's primary focus on being a realtime microkernel and introduced threading to the architecture. It also supported the by then more "modern" PC hardware, and of course ethernet. Neutrino came after that.