Now that I think about it, there is almost no cost involved in fact. It would be fairly easy for a company like Canon to identify in advance those perfect wafers and to use these for large chips, while using non perfect wafers for smaller chips. From then on it is just a matter of process control to make sure that at least some wafers are perfect.
This is not a genius idea. Intel does the same with their CPUs.
I am tempted to suggest that the testing for defects and such is done electronically on chips after they are fabricated, not by optical examination of blank wafers in advance. But what does either of us really know about the details?
What we do know is that several companies compete in the market for large sensors (most famously Kodak and Dalsa) and so if it really were as easy as you say, and if we avoid internet cynicism about industry leading companies being more ignorant thus us internet pundits about how to reduce costs or increase quality, or conspiracy theories about Kodak and Dalsa fixing prices and in the price strangling the DMF market to their mutual detriment, then one or more of these competitors would have done what you suggest, and reduced its costs substantially (they both make lots of smaller CCDs too, so your idea fits their situation), and then reduced their prices in order to gain market share, and so on.
The actual evidence of MF sensor competitive market pricing instead indicates that with the current state of the art, the costs of MF sensor design and fabrication is such that the price to customers for even a 44x33mm sensor is well over US$1000, and prices will scale at least in proportion to area. More according to every source I have read on chip yield.
And in case anyone suggest it, large CMOS sensors are not cheaper than equally large CCDs to make; that is an internet myth, based on falsely extrapolating the cost savings for tiny "camera on a chip" cell-phone and web-cam modules. Note well that the last remaining DSLR CCD's are clustered in the cheapest models!
Short version: I do not believe any internet forum proposal about dramatically improving price/performance beyond what competitive market forces are currently producing unless it is backed by evidence of a technological break-through that it too new to have been commercialized yet.