I think I will try to go the route of promoting adapter boards, with each ribbon cable being replaced with two or four inputs and same number of outputs and the other end extracting into 8 in/8 out or whatever that specific device may require.

That would decouple choice of voltages, selection of series resistances, pull-up resistors, EMI-filters, ... from the control electronics, while reducing problems with end-of-life for critical components, basically making the main processor the only really critical component. "Owning" both sides of the cable means that the cable signals can use optimal drive strength etc, before switching to correct voltage levels, current drive etc.

Of course, I'm not into the position to make such a decision. But it should be a presentable solution that does have advantages not covered by a single, huge, FPGA.