Distributed SQL sharding in YugabyteDB involves splitting tables into smaller units called tablets, which are then distributed across multiple servers to scale out horizontally. The ideal number and size of tablets depend on various factors such as data growth patterns, query activity, and storage constraints. Small static tables, frequently accessed with joins, can be broadcasted to a single tablet with duplicate covering indexes, while medium and large tables can be split automatically using hash or range sharding methods. Very large tables growing with business activity may require partitioning to optimize future archiving or purging operations, and huge tables should not grow indefinitely beyond the default splitting thresholds to avoid operational challenges. By understanding how sharding works and tuning the auto-splitting knobs, developers can ensure optimal performance and scalability for their distributed SQL databases.