If your pipeline is mostly PySpark/Scala, rename columns in the DataFrame to match the target and use df.write.saveAsTable. If your pipeline is mostly SQL (e.g., on SQL Warehouses), use INSERT … BY NAME from a temp view (or table).
Performance is broadly similar for both paths on large datasets. But it is just that the INSERT doesn’t handle schema evolution; for adding new columns, with pyspark way you get that benefit.