Yes, the OC potential/voltages required are a reality check for the more optimistic folks. Getting big clock speed increases on a new node is harder than ever.
This has little to do with process, but that the CPUs are way past the optimal frequency.
The side-effect(or advantage) of Intel being on 14nm and Skylake for so long is that Intel engineers were allowed to characterize the product for far longer than they have, or anyone else has been able to do in the past.
Before, at least the uarch or the process had to change every year. A year is a very short time in the development process. After taking out a few months to allow for inventory to build up, and coordinate with OEMs, you might have 6-7 months to launch the product. And all this time is spent on validation and finalizing the chip.
For 3 years, they kept squeezing out everything from the silicon every year. They said since 4790K that it can overclock to 5GHz, and it took 3 more chips to actually reach that(4790K, 6700K, 7700K, Coffeelake). Each time raising the ceiling by 100-200MHz.
5GHz is the frequency where you start making real sacrifices to get there because its a barrier. There was a speculation by Hiroshige Goto at PCWatch that Skylake may have been redesigned for higher frequencies. The intercore latency, the L2 and L3 latencies and bandwidth could all have been better if they aimed for lower frequencies, or had a replacement for Skylake arrived in 2016.