Yes, it is something unhealthy to seat on a bunch of far reaching ideas and still use artifact that even Microsoft tries to shake off (outsourcing their Office XML Ribbons). But as recent roar about monoids have shown - those who are Haskell most productive programmers are also most unlikely to tap on anything theoretical. Researchers on the other hand wouldn't be who they are if they thought mainly in terms of code. Perhaps some new model of cooperation between a researcher and a group of volunteers should be worked out.
Other work [4] seems to indicate that arrows are too general for the dataflow programming and co-monads are more suitable and hence might be usable for interactive interfaces.
I would be careful with moderating ambition here. It should be extremely open framework that could allow to accommodate all present experimental (mostly reactive) stuff but also allow close integration with functional database in the future. The GUI should also lower overheads for Geometric Algebra. I think that in the future GA, fractals and wavelets will be inbuilt on graphic cards and GUI might take a role of a sort of a synthesizer. But of course anything functional would open lots of new possibilities and give us a lot of fun. Cheers, Andrzej