Are there any recommendations when dealing with schema changes?

I'd say that the way described in https://github.com/jmriego/pipelinewise-target-bigquery is quite convenient (creating new nullable columns)