CREATE TABLE sandbox.orders_dev CLONE sales.orders makes a full copy of a table in seconds, whatever its size. Snowflake doesn't copy the data: the clone points at the same storage as the original. From then on they're independent. Change the clone and only the changed parts are stored again, so a fresh clone costs nothing in storage.
Clones work for schemas and whole databases too: CREATE SCHEMA sales_dev CLONE sales copies every table in it. That's how teams get a full-size development copy of production without paying to store it twice.
Combine cloning with time travel to copy a table as it was: CLONE sales.products AT(OFFSET => -60*30). It's a quick way to keep a known-good version before a risky change, or to recover one.