Structured Streaming provides native Kafka integration and processes incoming records in repeated micro-batches, making it suitable for the required sub-minute clickstream latency. For the customer data, an incremental Lakeflow Connect ingestion pattern tracks source progress and processes only new or changed records during subsequent executions. This avoids repeatedly transferring and processing unchanged Snowflake rows. Option B performs a complete extraction every four hours, so the later MERGE does not eliminate unnecessary source reading and network transfer. Option C relies on a time filter that may miss late changes, updates without a suitable timestamp, or records affected by clock boundaries. Option D explicitly performs full refreshes and therefore violates the requirement to prevent reprocessing unchanged customer data. Consequently, option A is the only approach that satisfies both the streaming-latency and incremental-ingestion requirements.

================