But really, is this such a big performance gain to say that if someone is not putting it he has forgotten about performance?
It depends on the application. If you run on this "case" some few time, no. If you may need to add thousands of elements or add into a new one very often, yes, you may see the difference.
Usually the most important optimizations come from design rather than tiny details of the .NET framework internals (which are important to know, I won't argue that, I wish all C# developers out there were forced to read CLR via C# and C# in Depth).
I agree. Until now, I never see a case where optimization can "fix" a design flaw. And, as Tridus said, maybe this is more a problem of knowledge. But somehow, I think it's part of developer responsibility to think more than just "do code that works", and maybe it's due the environment I'm working on.
Anyway, I'm very happy that Stardock people take care about performance early in the dev process, and not as other does, when the product is out and only if there is plenty people complaining about performance issues.