It would be 4 byte more per array for metadata, plus the sizes. It would actually require less memory than ragged arrays, which have to record the size in each

It would be 4 byte more per array for metadata, plus the sizes. It would actually require less memory than ragged arrays, which have to record the size in each subarray. API complexity could be handled by a polymorphism mechanism to make it possible to write code that is generic regarding row/column access order.