Current L2 utterance fluency literature tends to operationalize disfluency as isolated, individual features. However, disfluency features often co-occur at one location or across multiple locations in one utterance. This study explores the co-occurrence of L2 disfluency features in a speech corpus from 71 L1 and L2 speakers of English across proficiency levels on an elicited imitation task and an oral listen-to-summarize task. We segmented each participant's speech into analysis of speech (AS) units (k = 2,704), extracted 2,972 individual disfluency chains based on 15 disfluency variables, and subjected them to principal components analysis and hierarchical-based K-means clustering analysis to identify disfluency co-occurrence patterns and speaker profiles across tasks. Results showed that different disfluency co-occurrences can be interpreted around various repair behaviors, and these repair behaviors also differ across tasks. Further analysis concerning the disfluency--proficiency relationship suggests that whereas some disfluency co-occurrences are meaningfully associated with proficiency, others might not be.