This is a general comment. Over the years we've built and worked on many IDE, and richer they are with features, better adoption it gets.
The analogy of a notebook is somewhat similar to what a scripting developer would go through in perl/shell coding. For a statistician and scientist, this might not be much of a problem as their work isn't code heavy but more function heavy, with iterations, models, etc. But surely there is Java, Scala, Python that is being used here. And for a java developer, switching between JBuilder, IntelliJ, and then notebook is not quite consistent.
Not trying to make a developer out of a scientist, or vice versa, but trying to see how both worlds can meet and working on a notebook easier.