What are the advantages or maybe the disadvantages to have the .r files separated from the sources?
Is it a good general rule to separate them?
Separating them is to my opinion a good rule of thumb.
As an example, we do not give end-user site the source. So distribution of R code is easier and faster too.
Also, if delivering in many languages (eg French, English , Spanish...), we build a separate COMPILED Folder, with sub dir for each language (EN, FR, SP etc). Again its easier to deliver.
Also you may delete R code often, as they are usually less important, since you can recompile anything in few minutes. A must say, we have built a compiler procedure of own own, than can compile automatically the same app in up to 20 regional languages. A language being a Context more than an official country language. Ex: Terms CUSTOMER and CLIENT are the same word , but used in different markets,