I believe there are some gaps in test coverage here:
- Should be able to run a large NAAJ with spilling disabled (need to test both success and OOM cases)
- Test spilling cases:
- null_probe_rows_ does not fit in memory
- null_aware_probe_partition_ does not fit in memory along with other partitions
- null_aware_partition_ does not fit in memory along with other partitions
We should address these gaps before releasing the new buffer pool code to ensure we don't regress anything.