Clock speed scaling (Dennard Scaling) mostly stopped around 2007: https://en.wikipedia.org/wiki/Dennard_scaling That is not a new development.
Since then CPU performance developments have emphasized improving Instructions Per Clock cycle (IPC), adding additional cores, and adding specific hardware for certain cases (AVX instructions, Quick Sync, etc).
The ability to wring out additional IPC performance is dwindling, as most CPU architectural tricks have already been used.
For computers with Sandy Bridge or earlier CPUs, upgrading to Skylake may bring significant improvements. For computers with Haswell or Broadwell CPUs, the situation is less clear. E.g, the Haswell "Devil's Canyon" i7-4790K CPU was used in the 4Ghz 5k iMac. The fastest equivalent Skylake part is the i7-6700K, and it's only about 5% faster overall: http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation
Intel could put 8 cores on mainstream non-Xeon i7 CPUs, but it would be an expensive part and many iMac workloads would not benefit dramatically. It would greatly help aspects of FCP X, Lightroom, etc -- anybody who has monitored the activity of those with iStat Menus or similar utilities can see that. They are heavily multithreaded, frequently CPU bound on all available threads, so the more cores the better. However that does not characterize general use and for those specialized cases there's the Mac Pro.
