> CP/M did have a sort of revival in that it became common in low-end machines like the C-128

Amstrad were good late 80s CP/M machines. We got both those and C128 in New Zealand.

> the one RISC/CISC CPU thing that really mattered!

Not only indirect addressing, but also multiple memory operands in the same instruction — more than one VM page, really, though a single unaligned operand crossing a page boundary is also bad. Many machines trap on that case to this day and let software emulate it.

Not being able to easily tell how long an instruction is (and thus where the next one starts) is also bad, but can be overcome at some cost in the front end, and the back end is unaffected. Unlike x86 and VAX the 68k does actually tell you everything you need in the first 16 bits, but yes the complex addressing of the 020/030 were what killed it.