I agree- it would be a difficult decision to rewrite working code without a compelling reason. I could expand that point, but lets not go there for now. ;-) We could also observe that people in this forum considering NET are facing a rewrite anyway, so the opinions of those with "existing significant investment in older NET iterations" aren't so compelling.

FWIW I can report my own experience (anecdote) which is that C# code written 3 years ago is increasingly difficult to maintain because people seem reluctant to work with its older ways. They want to rewrite. That's the trouble- despite your dislike for dLinq (yes I know that today's label is "Linq to SQL" but that's a mouthful) there will be more and more developers relying on it or entities, with little or no experience with pre-strongly-typed-dataset technology. I think you're OK with your strongly typed datasets fwiw, but lets not go there now either. ;-) Can we agree that unless people are able to do it all themselves, the ability to support superseded methodology must be a significant driver.