I guess I see things differently.
For example, I am always willing to pay nearly double for a 4-layer PCB, when with a little extra effort a 2-layer PCB could be made to work. It increases routing flexibility, often helps signal quality, and it's nice to have proper power and ground planes. I think of going to 4-layer as just a component cost that's not even a big one. And it saves so much time during layout!
With this proposed future "super QL" the trick will be to offload as much of the vintage functionality as possible onto a single SuperIO device. This would require light OS/driver support if the SuperIO is chosen carefully so the constituent parts are mostly already used in a QL system already. For example, the PC87307 has the floppy controller used in the SGC, a standard DUART, standard PS2 keyboard/mouse, standard RTC.... This hugely reduces logic and pin costs on a CPLD or FPGA.
The sticky part is video. Peter could do the amazing job he did only because he kept the entire video and CPU section internal to the device. If he had to use external pins, it would have blown the budget.
If we look at the Q68 (and these comments are not questioning the business decisions or pricing of anyone involved with the Q68's production/sale) the price is £150. The component cost, rough estimate, is £25-30. If it used a double cost FPGA, would it really affect the final price that much? Most of the cost isn't in the parts, but in the development and actual manufacture labor. The components are only a tiny element of the final cost. The only way to lower the sale price is to buy components in bulk, to assemble in longer production runs and to sit on vast stocks for ages while they sell.
So yes, I'm 100% ready and willing to buy trays of expensive parts. And to use brute force approaches when the time and skill resources for more refined and efficient approaches just don't exist, meaning the project would take too long, be too costly or just never happen. It's a trap I've fallen into before.