Use case
MergeTree tables spend much time loading data parts sequentially. This happens for tables with granular partitioning (example: a table partitioned by day has data for ten years).