It was also common to invent "tiny languages" (e.g. PIC) and domain-specific languages (e.g. Awk, Yacc) to a greater extent than now. There was no real community to share work. I heard of OO and C++, but didn't really understand it, so when I tried to make my own I made functions that dispatched on the type of the first argument, then another version which could dispatch on the types of any the arguments. We would call this data-oriented now. I suspect that a lot of new stuff arose from such misunderstandings.
True, though I guess they were all working in more isolated ways than we are today too.
Probably the compilers were very much tied to the machines they were written on as well (I know we have lots of architectures today but I get the impression it wasn't like today where x86 dominates).