I think in order to create good software that fits the industry's current best practices then you do need a really in-depth domain-specific knowledge. If that's the goal then learn everything possible about that industry.

To create good solutions that are also innovative, I think that programmers only really need enough knowledge in that domain to be dangerous - to know more or less what happens but with enough gaps to not really know what's doable and what isn't. But, this knowledge is only useful if you can combine it with bits and pieces from several completely separate domains.

If innovation is the goal then I think there is a point where domain knowledge actually hurts you.

Of course there still needs to be at least one specialist to keep everything on track though - and I completely agree with you about having someone with deep knowledge on board. I think it should be someone who has been in that industry and will stay in that industry.

I can't see the point in programmers trying to become experts in areas that they're not interested in working in for the long haul. It's easier to learn (and actually apply) a little about a lot than a lot about a little.