The 3070 and 3080 has a large gap in performance, so it might be better to release a SKU that is 15% faster than the 3070 with 12gn ram or more, probably the rumoured 103 die. But even then, having a 3070ti with more ram than the 3080 is a bit of a strange situation.
I don't know if they're accurate at all, but the alleged scores for the 3070 put the 3080 about 25% above it, but that's just an artificial benchmark and the numbers could be fake anyhow.
The other problem is that even if there's a much larger gap than rumored, NVidia doesn't have any silicon to make a 3070 Ti that would be 15% better than a 3070 without making a new die since the 3070 is nearly a full GA-104 die.
If the 3070 really is that close to the 3080 it's a great deal at $500. NVidia doesn't need a GA-103 die, they just need to make a lot of 3070 cards. Since it isn't using the newer GDDR6X it should be easy for them to make a 16 GB version at some point if they really wanted to. It would also mean that more cores don't scale well for Ampere since a 3080 had over 40% more cores and memory bandwidth than the 3070. Why make another die to slot between them when it won't gain all that much if the 3070 is at the point of diminishing returns.
However, if the 3080 is a lot further ahead of the 3070, is there enough price room for a GA-103 die that has to come in between $500 and $700? It would likely have the same problem as the 3070 at that point where a 3080 is just a better value / dollar. Maybe it makes sense in the long run if NVidia needs a refresh. If they make a GA-103 that could be sold uncut as a 3080 SUPER and slightly cut as a 3070 SUPER then they have a good reason for it, but that seems like something for next fall or late summer at the soonest. By then there should be some 2 GB GDDR6X modules that would make another compelling reason for those products to exist.