I would say, translating comedy. Namely — wordplay, puns and cultural references. This is because audiovisual translation operates not in one channel, as the regular one does (written word), but in four — spoken language, sounds (including music), picture (=video) and on-screen text. This complicates matters, well, fourfold, because you are constricted by four planes of meaning, and not just one.
For example, imagine translating the idiom "between a rock and a hard place" in a book. You just find a similar expression in the target language and go with it — or, if it does not exist, you can translate descriptively or coin something new. Now imagine you are translating the same idiom in a comedy film, and the video sequence shows a person literally standing between a rock and some hard place, maybe even with a sign plate saying "Hard Place". What do you do now? If you find a similar expression, chances are, the video won't make much sense, because in the target language the idiom has nothing to do with both rock and hard place, which is the case in Russian. Now you have to come up with something else and think much harder than you normally would translating a book.
And this is why translating comedy is, in my opinion, the most challenging aspect of subtitling.
On the technical side of things, working around shot changes is quite difficult. Along with the time (reading speed) and space (row length limit) constraints, it restricts your choices of how to segment the script into subtitles.