So, I guess the question is really a throw-back to the 1060 era then. That is to say whether there was a niche for the 1060 3GB versus the 1060 6GB. Hmm....when framed as such I'd gladly buy a 3050 or 4050 with only 4 GB. My choice was only if there was no other option...which is the current market.
Of course, this then needs to be framed with a cost to value proposition. If a base level card is 200 USD, and doubling the RAM is only 240 USD, then there's literally no reason to purchase the 4 GB thing. If it's a matter of a 100 USD base price and a 120 USD price (the same % difference) then the math is very much different. That is to say when the delta between the two is so much smaller and the cost is lower then there's a whole lot more reason to jump up a model to get extra VRAM. When you've got such a huge investment off the bat, then the lesser VRAM options make no sense.
I know this is counter intuitive, but it's a permutation of the sunk cost fallacy. If I'm looking at a lesser investment then, saving a bit more for less VRAM is tolerable because I can go up a level if I wanted to. If I have to spend twice as much base, then the extra investment is "just a bit more." If going from a 3050 to a 3060 is 200-400, and the space in-between is a 3050 with 8GB of VRAM for only 240, then it makes less sense to buy a 4GB model.
Now, let's talk pre-coof. The 1060 offered a 3 and 6 GB variant. The delta from these was about a 20USD, but jumping from the 1050 to 1060 was about 60USD (at one point). I bought the 1060 3GB model because it significantly out performed the1050, but the cost to double the VRAM made precious little sense when very few programs and resources used it. Hopefully this is the case with the 40x0 series.. but right now it just doesn't make sense to plunk down the crazy money for a discrete GPU, but to then cheap out on VRAM.