In my experience teaching R, the single biggest issue is that it is a dynamic language that is in love with silent casting [or really returning objects with unexpected types]. Although other bugs might be more common, none will frustrate the user more than the type of a variable changing without warning, and most of the time the error message refers to something else.

A couple of examples are stripping the time series attributes of a ts object and default conversion of a row of a matrix to a vector. These are the cases where they come to my office and say they have no idea what's going on. After using R for a decade, I know the language well enough that these are about the only errors I get.