pg_repack has some significant downsides in its implementation; I question whether it’s really a default over re-indexing concurrently. I’ve certainly not gotten that impression, and we maintain many very large Postgres clusters.