Currently, the behavior of the resharding coordinator is sensitive to the cluster’s active FCV. This sensitivity makes it harder to ensure correctness during rolling upgrades or downgrades. To address this, the resharding coordinator should maintain a fixed and stable view of the FCV throughout the entire lifetime of a resharding operation.

Adopt the same pattern CAR uses for long-lived DDL coordinators: capture and persist a VersionContext at coordinator initialization and then use it to fix the coordinator’s view of FCV for the duration of the operation, regardless of subsequent FCV changes.