It can be useful for iceberg.
Also, in https://github.com/source-cooperative/csv-table/, I'm parsing a CSV. I want to start by fetching only the first 4MB (for example) to show the first rows, then fetch more data if needed.
We need to know the number of rows to be able to compute the scrollbar size. As we cannot easily know the exact number of rows, we could use an estimate (total_byte_length * init_num_rows / init_byte_length) for that purpose.