Mirroring time-series CRUD operations into a regular collection allows us to validate the data being stored more heavily. We have a targeted concurrency workload doing something like this already for inserts, but having it done across all tests and other operations would give us higher correctness confidence.