you could use the Group Loop Start node to do things per group. This will provide you with a flow variable that contains the current value of your group you could use to create a CSV file name.
As a result, your data will be split by the group you have defined.
If you need to split the data further into chunks of a special width (500 lines per CSV and then the rest of each group into the last file). You might take some inspiration from this workflow: