Live broadcast simultaneous interpreting on television is widely acknowledged as one of the most challenging and stressful forms of screen translation, and translational activity in general. Aside form experience-based accounts by media interpreters describing the specific working conditions and constraint in TV interpreting, the literature includes some interpreter and user surveys as well as case studies of physiological stress, but very few corpus-based analyses of actual media interpreting output. Based on a substantial corpus (approx. 64,0000 words) of three comparable sets of English-German TV interpreting of a US Presidential Debate, this paper examines how highly professional media interpreters cope with such complex source-text material as cultural references in simultaneous interpreting. Framed by a functionalist account of the communicative event and the professional assignment, the corpus-based analysis presents quantitative evidence of how the interpreters cope (or not) as well as qualitative date illustrating particular problem-solving strategies.