If 5ns memory is sold as PC2700, of course that would be a good thing for us. It would run at that spec just fine. Maybe you read that some manufacturers use better than 6ns chips so they are really up to spec to not run into problems.
An example for this would be Kingmax PC2700. We all believe to know it has 5ns chips, but it's not quite true. Those are actually 5.8ns chips, rounding to 6ns would be correct but would underspec it, so they just put -05 on the chip and let it be. They use 5.8ns chips to be truely up to the PC2700 spec. If I say they use 5.8ns, they don't have a machine that they can enter the value they want the chips to come out, reality is far from that.
The wavers they get the chips from have pretty high quality fluctuations. It can happen only by a marginal change of environment during production/manufacturing. What they get is normally 7.5ns chips because they are taking the worst case scenario. From our info, pure 6ns production is not made anywhere on the world as of today. The only way they get the better chips now is by testing them. For the example of kingmax, whatever chip manufacturer is producing the chips starts testing a defined quantity of 5.8ns chips for kingmax, they test chips until they get the right quantity. Maybe some get sold as faster chips as well. The rest gets sold as 7.5ns chips or whatever the "worst case" it was that they set.
In the same context, one interesting thing is Corsair XMS2700. By now everybody knows this memory has "overclocked" 7.5ns micron chips. If you look at the chip, you'll see it's 7.5ns written on it. Alot of people were disappointed because this memory is about the most expensive out there and has "only" 7.5ns chips. However, this is not the whole story. These chips actually are up to spec without overclocking, meaning indeed they could be talked of as 6ns chips. What happens here is that Corsair, as opposed to other manufactures, does all the testing in-house. So unlike the above example, they buy the untested chips from micron and then start to test them by themselves, that being the only difference. This explains why the chips are printed as 7.5ns chips already, but are really the chips that we all want: tested for 166 like all others. While Corsair is the only PC2700 specced for CAS 2. If you read out the eprom info on this ram, most likely you will get an indication somewhere near 6ns. Some people were offended by that because they thought Corsair were cheating, thinking that the printed value would be right, and not the programmed. Funnily, more likely it's the other way round
You are saying that 5ns is optimal for PC2700, and yes you are right. So is 4ns and 3ns if you want, you at least have to use exactly 6ns or better chips.
From the math, 6ns is exactly 166 Mhz.
Hope I could clear some things up.