In database normalization theory, which anomaly specifically refers to the unintended loss of information when a record is removed, because that record happened to be the only one holding a particular fact?

A deletion anomaly occurs in a poorly normalized table when deleting a row removes not just the intended record, but also destroys some other piece of information that happened to only be stored as a side effect of that same row — for example, deleting the last order placed by a supplier might also permanently erase that supplier's contact details, if supplier details weren't stored in their own separate table. An insertion anomaly instead prevents recording a fact until unrelated data is also available; an update anomaly forces the same fact to be changed in multiple places; 'concurrency anomaly' and 'referential anomaly' as worded here describe real but different classes of problems, not this specific loss-on-deletion scenario.