FTR I am not sure I believe that explanation. I was with Gavin a few years ago when Andrey Breslav asked us if we were going to implement reified generics (we had not yet at the time) because they were having trouble implementing it and so if we were not going to implement it, they would not bother.
We haven't talked about this conversation publicly in the past because to be brutally honest most of it would make them look really bad, and it's likely that we won't, but this anecdote was very relevant to the discussion about reified generics.
My guess is they tried and failed, strictly based on this conversation. It's possible that they really considered it too expensive, but I since that's not backed by public experiments and our own experiments tell us it's not that expensive for the benefits it gives us, I don't _have_ to believe them ;)