The scenario is as follows:
- A query joins a large partitioned table with another table on a non-partition column.
- All of the rows on the build side are filtered out, so the join has an empty build.
Currently what happens is that all partitions of the large table are scanned and the runtime filter filters all the rows out after materialization.
What would ideally happen is that the scan would notice that the runtime filter always returns false and prune all of the partitions.