IGNORE_DUP_KEY — what it is, since when, and the performance trap

An old SQL Server index option that drops duplicate rows on insert instead of failing the whole statement, handy for idempotency but ~18x slower on a clustered index.