If Intel has been unable to make significant improvements to 10nm yields under the 10nm+ process node; then Intel is in trouble period. Maybe this is why Ashraf Eassa thinks that Intel should be planning a fourth 10nm node. Maybe that node should use EUV, since it seems that SAQP is killing Intel.
Intel definitely seems to have had an issue with moving aggressively to new lithography techniques.
Remember with 22nm, which is probably the last "drama-free" node from Intel, Mark Bohr was talking about how the min metal pitch of 80nm was chosen specifically because it allowed Intel to avoid the hassle of double patterning the densest metal layers.
It was when Intel got it into their minds that they needed to leap ahead and "beat" the rest of the industry to new denser nodes, they started being much more aggressive with the lithography -- SADP at 14nm and SAQP at 10nm.
Had this worked out (and I suspect Intel had a lot of confidence in its ability to get this done), then they could very well have been as far ahead of the rest of the industry that they say they are. But it clearly didn't.
Intel may have been better served by iterating less aggressively with each generation but being able to make more iterations within a certain span of time.
Anyway, whatever Intel is doing just isn't working properly and it's ultimately hurting its ability to release cool new products on time. The moves they did with KBL/KBL Refresh/CFL are certainly nice, but with good planning ahead of time, Intel could've done better in terms of advancing its architectures and processes.