Having pulled it out, clean it up and make it work consistently with the new interface model. Then, put it back in as an available option. It is still a very powerful and viable rendering technique, and there are a helluva lot of still-important [commercial …] Blender projects out there which use it and still need it.
I think you are really underestimating the difficulty and I think this is a byproduct of developers making progress look so easy. Developing software on the level of Blender is REALLY hard work. While it looks like you can just pull bits out, polish them up and put them back, the reality is nothing like that at all and the cons outweigh the pros by a long shot. I would be guessing it would be weeks if not months of developer work and then it would need to be maintained.
Part of a good product is deciding both what to do and what not do do. While I did like to tinker around with BI, it’s days are numbered and the industry has moved on from renderers like that. It is just not worth the effort.
Unless someone has big piles of money to spend on converting it to an external renderer, it should be left alone in 2.79 where it is perfectly useble. 2.8+ has more important goals to accomplish like a materials library, advanced texture painting, opensubdiv and other features to make Blender stand out.
I would love to have older features to stay (Carve boolean removal makes me cry) but things have to move on.