When a bucket gets archived, it minimizes the memory footprint and only retains the necessary information to efficiently identify it as a candidate bucket for future insertions. The compression state is not necessary for this identification.