Index Skip Scan
An access path that can use a composite index even when its leading column is absent from the predicate by treating leading-column values as logical subranges.
With a composite index on (A, B), a predicate only on B does not satisfy the usual leading-prefix pattern for a normal range scan. A skip scan can, in suitable distributions, treat distinct A values as logical partitions and probe B within each one.
Its cost depends heavily on leading-column cardinality and predicate selectivity. A low-cardinality prefix can make the technique useful, while many repeated subscans can be more expensive than a table scan. The engineering meaning of indexes in relational databases explains why the access path must be validated against the measured workload.