Its a 45nm process, thats just about the maximum its can take most likely. It just means that carefull choice of memory will be important, and there's most likely options to change the multiplier on the memory when you overclock. Its probably one of the the reasons intel went for a tripple channel memory controller. It's got a lot more bandwidth at lower frequencies. Intel processors are pretty good at prefetching data to avoid bandwidth bottlenecks, and even with the slowest ddr3 is going to have a lot more bandwidth than Core 2 had.
Timings may be important, but dont get too caught up on the, after all 10-10-10-24 @ 1600mhz is the same latency in time as 5-5-5-12 @ 800mhz, or 2.5-2.5-2.5-6@400mhz. i7 will have reduced latency compared to core 2 because its got rid of the external memory controller.
You cant compare it to an opteron, plenty of people blew P4 Northwoods, because they were used to putting 2.0V into their 180nm processors and the 130nm's were more sensitive. Everytime the processes is shrunk, the chips get more sensitive to over voltage.
The only difference here, is both CPU voltage, and Memory voltage can both fry the CPU, because the memory controller is on the cpu die. Im sure the memory manufacturers of "overclockers" memory will qualify new batches of memory as "nehalem/i7" approved for overclocking at low voltages.
The CPU itself should be plenty overclockable, and the people cooling with air should have no worries, just be sensible. Water and Phase overclockers, well they will have to work out whats best for them. Most likely high CPU clocks, but not pushing the memory too far.