D happened at the wrong time. It was a better C++ that nobody really asked for at a time when java and C# were the future. When some years later that surprising wave of alternative languages for CLR and the JVM arrived and opened the door wide for other languages as well, D was already stigmatized as the better C++ that did not make it.

I also think that, due to it's early start, D simply fails to appeal to non-C++-developers. The other new native compiled languages, Go and Rust, are not just a new toy for C++ people, but are also seen as long awaited "on metal" opportunities for people from other language backgrounds (python and ruby for go, type-heavy managed languages for rust), who would never touch c++ or, guilty by association, D. This is a much bigger audience than the subset of c++ devs willing to give up their hard-earned mastery of c++. The perceived promise of native for high-level devs might not hold, but this won't help D at all.