Modern code review tools, such as Gerrit and GitHub, play a central role in coordinating large-scale software development activities. In Gerrit, developers can associate related changes under a shared "topic'', thereby indicating that these changes collectively form a coherent logical contribution. Although topics are widely used in large ecosystems such as OpenStack, little is known about how they are actually structured, managed, and coordinated in practice. In this paper, we present a large-scale empirical study of the ``topic'' feature in the OpenStack Gerrit instance, covering more than 900K change, 2,372 repositories, and 336,111. Our study first explores the internal structure and prevalence of topics to understand how developers organize related changes. We then examine how topic-linked changes differ from standalone reviews in terms of review dynamics and development activity. Finally, we analyze the sequencing of changes within topics, investigating whether the order in which changes are merged aligns with their original submission order. The results show that the topics are predominantly small and localized: 76% contain a single code change and more than 93% are confined to a single repository and branch. Compared to standalone code changes, topic-linked changes undergo significantly more revisions, have longer review durations, and trigger more discussion messages, inline comments, and CI/CD jobs, indicating higher coordination and validation activity. Analysis of merging behavior further reveals that a large fraction of topics (approximately 40%) are merged out of submission order once they exceed six changes. Topics merged out of order exhibit more revisions, reviewer interactions, and CI executions, whereas those merged in order take longer to complete, but involve fewer interactions. These findings highlight two distinct integration modes in Gerrit: sequential integration, which is slower but more structured, and parallel integration, which is faster but coordination-intensive. Overall, our study provides an empirical characterization of Gerrit topics as a lightweight yet coordination-rich mechanism, offering practical insights for improving code review dependency management and merge-order support in large-scale code review systems.