Music records are largely a byproduct of collaborative efforts. Understanding how musicians collaborate to create records provides a step to understand the social production of music. This work leverages recent methods from trajectory mining to investigate how musicians have collaborated over time to record albums. Our case study analyzes data from the Discogs.com database from the Jazz domain. Our analysis examines how to explore the latent structure of collaboration between leading artists or bands and instrumentists over time. Moreover, we leverage the latent structure of our dataset to perform large-scale quantitative analyses of typical collaboration dynamics in different artist communities.