Well, the Zen 2 CPU has hyperthreading, unlike the A78C, so the number of threads is the same, and the A78C cores are expected to be clocked lower. Without an apples-to-apples comparison of the two CPUs at mobile power levels, I don't think we can know whether 4 x Zen 2 cores is more efficient than 8 x A78C cores. But it makes sense to me to include more cores to try to compensate for a lack of hyperthreading and lower clocks.
T239 is seemingly too big for 8nm at mobile clocks, as there is a minimum voltage that the chip will support, and decreasing clocks further will just result in throwing away performance for no gain in efficiency. If it was really 1024 cores, it would make a lot more sense at 8nm, and that's what everyone expected. However, it can't be cut down, based on the Nvidia leak, because the API was showing that all 1536 cores were available.
That's exactly why people think it is using a newer Samsung process, with reduced power consumption. Is that unreasonable from a cost perspective? Well, The APU in Steam Deck (Van Gogh) is 163mm^2, and T239 is estimated to be 200mm^2, or 25% bigger. (We would genererally expect Nvidia would to demand a greater margin than AMD, but by the same token the Switch 2 will be a much higher volume product).
However, if Nintendo are using a newer process, we would expect that Samsung has given them/Nvidia a very good deal, to make this worthwhile.