When a sharded timeseries collection is created in legacy format, its config.collections entry does not include the timeseriesBucketsMayHaveMixedSchemaData field, since that field is only written for collections created as viewless.

When the FCV is upgraded and the legacy collection is converted to viewless format, the timeseriesUpgradeDowngradeCoordinator copies the existing config.collections entry as-is, carrying over the absence of the field.

This means the upgraded collection lacks a field that freshly-created viewless collections always have, leading to an inconsistent state in the global catalog after FCV upgrade.

Conversely, on downgrade from viewless to legacy, the field must be removed since it is not part of the legacy format.