This would allow us to compute the size of DataCells (a very costly operation) on the clients and therefore:

• distribute the load to the clients rather than submerging the server
• do the computation on the batching cold path, where we have all the time in the world