Fly-by-wire works, it’s reliable, and it’s here to stay. We haven’t had stick-and-rudder guys for decades on commercial flights.
What problems or dangers would be posed by adding a minimalist mode to the software? Granted, one has to make sure that it doesn't get invoked accidentally, but that shouldn't be difficult.
I would expect that a fairly minimal control system for an airplane could be handled with something fairly primitive on the level of an 8x52 microcontroller or equivalent. The system wouldn't do a whole lot--just direct control surfaces to match the positions of the pilot's controls. Communication wouldn't have to be terribly fast, so the backup communications bus could use clunky but noise-resistant protocols.
The only per-unit cost I would think would be at all significant would be stringing cables for the backup network. The software would be much simpler than that which would normally be used in flying the plane, so I would think the extra cost of certifying it would be slight compared with the cost of certifying the main software.
“but that shouldn’t be difficult.”
maybe, maybe not.
What IS difficult is making sure pilots know how to fly a plane with completely different flying qualities than the one they took off with.
Why go minimalist? why not have another fully-redundant flight control system?
triple, quad redundant how much more do you need?