A column changing its data type is generally considering a breaking change for the schema (for obvious reasons), while adding more columns isn’t. Backwards-compatible schema evolution isn’t practical without the latter — you’d have to add a new secondary table whenever you want to add more columns.
This mirrors how adding additional fields to an object type in a programming language usually isn’t considered a breaking change, but changing the type of an existing field is.
A column changing its data type is generally considering a breaking change for the schema (for obvious reasons), while adding more columns isn’t. Backwards-compatible schema evolution isn’t practical without the latter — you’d have to add a new secondary table whenever you want to add more columns.
This mirrors how adding additional fields to an object type in a programming language usually isn’t considered a breaking change, but changing the type of an existing field is.