We have a large table in Snowflake that's around 700 million rows that loads from a parquet file. Occasionally there's a schema change, and Fivetran tries to update the data type on the table.

1. Create a temporary column with the updated data type
2. Update the temporary column with the type-cast values from the old column
3. Drop the old column
4. Rename the temporary column to the correct name

The problem is that our regular Fivetran warehouse isn't large enough to update all 700 million rows at once. We have fixed this by manually running the above process, but for step 2 we update in batches - usually based on _fivetran_synced. 

Request for Fivetran: If a query like step 2 above times out, batch out the update statements to contain the impact on the warehouse and hopefully get the query to run without timing out.