This is a different problem. Update is expensive in distributed columnar data. ReplacingMergeTree translates updates into inserts which are very fast and always

This is a different problem. Update is expensive in distributed columnar data. ReplacingMergeTree translates updates into inserts which are very fast and always have been. It then updates rows in a lazy fashion.