Scientists in the natural histories create the frameworks, calendars and infrastructures that allow us to understand and grapple with "deep time" -- but they do so within their own temporally complex scholarly settings: they draw on classification systems that are constantly facing revision and methodological revolution; database systems that simultaneously face forced obsolescence and true decay; and data collections in need of maintenance and migration. In this talk, I consider the rhythms of fracture and reconciliation in the data infrastructure in the natural sciences. This talk draws on my on-going work studying long-lived data infrastructures in memory institutions through the "Migrating Research Data Collections" project, as well as prior work studying taxonomic data practices through interface design.  Some of this work has been previously published (dl.acm.org… ; doi.org…);  this talk presents emergent findings from new work, and synthesis of this prior work.