Regarding the mistake of creating extremely wide tables (100+ columns), as long as you don't include all of those columns in a query, Spark shouldn't run into the "spark.sql.codegen.maxFields" limit.
