It might be the case and it might be highly dependent on how adding DLSS fits into the general renderer architecture. However, you’re also making an assumption here, which is scaling with DLSS or scaling with the CAS Shader (what’s in use right now) is in any case taking the same amount of resources.
In effect, when using DLSS you’re actually using dedicated silicon (transistors) which are doing nothing otherwise, whereas when using the CAS Shader, you’re using the same CU you’re using for rendering.
DLSS is not just “scaling” like the CAS Shader, it is “creating” new pixels which didn’t exist before so that it looks like it was natively rendered at a higher resolution and this is done with dedicated silicon which is not contributing to the rest of the rendering. You’re really adding up both visuals and performance in using DLSS.
Now I agree with you TAA70 in 4K is already really good looking, but in VR it is catastrophic with EFIS because FS2020 is rendering EFIS based on the game render resolution (in your example 70% of 4K) not the post-processing resolution (in your example 100% of 4K) and this shows a lot:
[BUG/FEATURE] EFIS Screens Problems and Solutions for higher legibility