First, these systems were "obvious" and could be understood in minutes from reading a manual. Second, there were very few different kinds of computers, and the community of system programmers was similarly small. At least in the United States, almost everyone who wanted to know about these systems could and did communicate directly with their authors.
One gets the sense that few anticipated there would one day be a computer in every home ;)
Anyhow, back home, our GEC 4000s¹ were fast and pretty reliable, running the OS4000 operating system. I've never confirmed it, but I assume "4000" was a tribute to RC 4000, given its similar software design and terminology, though Nucleus, (the moral equivalent of?) the microkernel, was in hardware/firmware. It's an interesting counterpoint to the RC 4000 monitor generally described as slow. As the GEC 4000 architecture was later emulated decently fast on M88000s, the speed wasn't just down to the original hardware; such 1970s microkernel-ish designs could work well. Skål PBH.
After Karger et al pentested it...
...it also got a security upgrade. Its security profile was stronger than most other OS's. Although mostly a flop, its combo of features and security did cause some organizations to buy it. The last one shut down in year 2000. It was just too expensive ($7+ mil?) to justify in a market that, at the time, mostly cared about performance per dollar with security and maintenance activities a non-concern. On other hand, Burroughs B5000 and System/38 (later AS/400) were reliability/maintainability-focused products from that era that did manage to survive to present day.
What I liked to see in updates are secure and fast microkernels (L3,L4) and secure nonblocking parallelism (Singularity) in a secure language.