When generating schema synchronization script, add new columns (which are existing in the Source object and are missing in the Target object) before dropping old ones (which are existing only in the Target object).

This will preserve data integrity.