Thus a GATHER_SCHEMA_STATS call can be made to have Table Statistics gathered on multiple tables concurrently.
In earlier versions, the DEGREE (of parallelism) would only apply to large tables. Thus, in a schema of some thousands of tables, the GATHER_SCHEMA_STATS call would run through all the tables serially. Only at some large tables would it use Parallelism.
The introduction of the CONCURRENT preference allows the use of Parallel operators to work on multiple tables concurrently.
The Optimizer team has released a short write-up on this.