CD0-CD7 should be color data of graphics. This requires a logical analysis. If the palette problem is corrected, it will be of great value
Well, sure. I was curious about the specifics.
Maybe you can put some CPLD logic between UM6558/UM6559 on CD0-7 lines, which changes problematic dot colours in the fly to the correct ones, that don't cause troubles?
It looks like it's only 8 bits wide input and output ... I bet you could just use a fast ROM. After all, we can easily get 70ns ROMs, and it's a ~1/(186ns) pixel clock...