Because they're already making T239. If you were to alter Orin to make the best Shield you could, you'd get something close to T239 anyway. Nvidia doesn't have to design a new chip, which is cheaper for them, and Nintendo doesn't pay for exclusivity on the hardware, which is cheaper for them.
There are only 2 real places where T239 isn't a good fit. One, it's got custom decompression hardware, and two it's kinda beefy. In theory, both of these "problems" can be solved by binning. If some of the chips coming out of the foundry "work" but don't have all the hardware functional, they ordinarily can't be used, but if Shield doesn't need all that hardware, then potentially you get a smaller T239 variant (with, say, 8SMs instead of 12) for "free".
If Nintendo buys a wafer to make T239 for the NuSwitch, and some of the chips can't be used, that's lost money for silicon. But if some of those binned chips can be used for the Shield, Nvidia pays for those. That recoups Nintendo money for what would usually be waste, and Nvidia only has to pay for the chips that actually meet their requirements. Once again, it makes it cheaper for everyone.
The Apple TV 4k is made from binned chips for the iPhone 13/14. Having multiple product lines for big chips like this is a smart move, and it's clear Nintendo doesn't view the Shield products as competitors. What's more interesting to me will be if Nvidia decides to market T239/T239b as a "non-automotive" Orin. The automotive industry has been the biggest consistent Tegra customer, but previous Tegras all wound up in tablets/laptops/AR devices. It's possible that the chip itself will be Nvidia's second business line, rather than a Shield. After all, Nintendo will be proving the chips value, Nvidia doesn't need a first party product to convince integrators to use it.
I have been looking but I haven't seen anything new. Do you mind linking me to what you've found?