For assessment to be equitable, it is central that teachers/raters perceive and apply grade criteria similarly. However, in assessing L2 oral proficiency in paired tests, raters must grade test-takers individually on a joint interaction performance. With a conversation analytic approach, we examine closely one recording from a 9th-grade national test of L2 English with an aim to uncover some aspects that underpin vastly divergent assessments (as assigned by three raters) of one test-taker. Findings pointing to issues such as moral stance, rater experiences, and the interlocutor effect are discussed in light of equity in L2 oral proficiency testing and assessment.