The question asks for two Snowflake keywords that enable querying data without performing a full table scan. Based on Snowflake documentation and the community discussion consensus (with DE receiving 80% of votes and detailed reasoning), SAMPLE and TABLESAMPLE are synonymous clauses that perform probabilistic or fixed-size sampling before reading all micro-partitions, allowing Snowflake to skip large portions of data. In contrast, TOP [n], LIMIT, and FETCH all require a complete table scan first before applying the row limitation, making them unsuitable for avoiding full scans. The community discussion highlights that SAMPLE/TABLESAMPLE are specifically designed for efficient sampling without full table access, while the other options operate on the full result set after scanning.