How different is API consumption between loader operations and bulk API operations?
Vault loader always uses batch API to load data, 500 data per chunk. So all the loader operations count for API limits. The number of API Calls that Vault loader makes depends on the data volume when loading or extracting.
For a Load with an input CSV that contains 10,000 rows, there are 20 requests which consume 20 API limits.
For Extract, it uses VQL and sends subsequent requests if the return data is more than one page.
If the return result total is 10,000 and if each page contains 500 items, there are 20 requests in total which consume 20 API limits as well.
In addition, there are some requests to retrieve metadata, such as checking header columns, but it is just one request before the whole load starts.