GitLab

The reason for this is that BLOCK_RECORD format is not good when there is a lot of duplicated keys as it first writes the data (to get the row position) and
then writes the key (and thus checks for duplicates).