this is a dx12 issue
as jayz said on dx11 3-way and 4-way sli are not affected
developers wanted MORE control over SLi - hence MDA mode and LDA modes
Its the Software developers that wanted this so they are responsible for this - and so should they be- since SLi basically WAS a developer/OS optimization issue
now they have more control over methods and optimizations - Nvidia only has to provide 2 way optimized ...most users don't go beyond 2 cards(sli) and maybe physx
most games don't scale well beyond 2 way anyways
you also have microsoft as a contributing factor - dx12 - win 10 - another person/company for developers to figure out how to handle 3-way and 4 way optimizations with
more people more issues
now the developers can determine they own way to optimize dx12 Sli above 2 way - without Nvidia and with Microsoft only
remember dx12 brought new features like unified memory
gets the extra person/company out of the room...easier to make progress
furthermore CPU bottlenecks are the issues coming up - gpu tech is overwhelming cpu tech and are hitting bottle necks at the cpu
and it's only gonna get worse
as jay said his 5960x was bottlenecked