YODA and the Excel format are both unfriendly to large datasets, Excel because of row limits, and Yoda because it appears to require loading the entire dataset into memory before being able to generate the YODA file. Supporting a CSV-focused packaging approach such as http://frictionlessdata.io/data-packages/ could make for a large-data-friendly ODM2 data format.