I investigate a case where two different methods of deriving an inline constant can lead to very different cardinality estimates.