It’s a tidy articulation of a problem I’ve also observed. It’s interesting how each party typically views it from the discipline in which they initially train and specialise. I assume most people fit in to this data scientist group as someone with a string competence in one area who subsequently bolts on extra skills when they are frustrated or limited by other factors.
For my part, my background was very much domain analysis (in your terms, I think we called ourselves insight analysts). Coming from that background, every piece of analysis/design/coding should have a link to a tangible business improvement which is predicted and measured. I’ve gradually expanded my skills with SQL dev, analytics software, Python, R. All the time, these things have to make it easier for me to deliver a business objective.
As long as you have a team of people who want to learn more about areas they don’t already specialise in which can support their specialisation, you should have a functioning unit. For every domain only expert, I’ve seen statisticians who make intricate and time consuming solutions to probalems which don’t affect the business much if they succeed or fail and engineers who make a hugely scalable platform for a product which is going to be shelved.