Thanks for the writer Chris,
I reached the conclusion that the easiest solution would be not to mess with the tufte class. Its main advantage seems to be the rational use of space. Tufte's design allows for packing information in a clever way, which is to say that whenever tables and figures exists in the document, they will take less space, whereas the amount of words per page is about the same of a tall (regular) latex article page. In my non-scientific tests, the difference was of about 15% space saved in heavily image or table packed documents. The difference is not big, if that would be the sole reason to use that design. I totally see why the R-team is insisting on that, as it is actually super useful to present graphs and tables.
Why would one need to save space? Some organizations in certain circumstances require a limited number of pages (talking about funders, sponsorships, which are the type of org that I deal with), sometimes the 15% more material can help you convey your findings better. The confort of reading a pdf in tufte's design is about the same as a well rendered latex document, as it keeps the number of words per line in the 60-80 mark depending on the font size.
Another strenght of tufte's design is that fullwidth figures allows for complex images to take more space than the actual text. The reader will be presented with a larger image to inspect, that could help in seeing the details, and so on.
So in the end, the true advantage of tufte's design is not in the fonts, or the fancy headings, or titling. The strength is on the larger margin where you can put notes and figure captions; and in the fullwidth environment.
These might be easy to achieve w/o even messing with the document class; floatrow package seems a great candidate to help achieving the aim of a "tuftesque" document in the sense described above. I will investigate whether it can mimic some of the design aspects mentioned, and started by
this question at SO.