SERVER-89769 added a majority write concern with journaling here and here to ensure the oldest stable timestamp is advanced, however, this is a wrong assumption on how the replication layer works, as pointed out in this bf, so, we need to manually wait for the oldest time to advance in the test before forcing the checkpoint with the fsync command here and here. We could do this, for example, by waiting for the following log line:

d20290| 2024-10-28T23:05:25.427+01:00 I  WTTS     22430       [JournalFlusher] "WiredTiger message","attr":{"message":{"ts_sec":1730153125,"ts_usec":426297,"thread":"12560:140733228867216","session_name":"WT_CONNECTION.set_timestamp","category":"WT_VERB_TIMESTAMP","category_id":40,"verbose_level":"DEBUG_1","verbose_level_id":1,"msg":"Timestamp (1730153125, 1): Updated global oldest timestamp"}}

This way we will make sure the collection will be reaped.