For queries like:
select count(*) from t where timestamp_column < xxx and timestamp_column > yyy, so far the filter condition cannot be pushed down to DT scan, thus introduces extra computation consuming more CPU.

The reason why we chose not to push down timestamp expression is that we need to deal with timezone. However it is actually possible to store min/max of DT in UTC, and transform them according to the required timezone (specified in DAG request) and do the comparison.