I can now process 100,000 GBIF records in about 20 seconds. I tried processing a million at a time but it was less efficient, so batches of 100,000 seem to be about the best performance. Later today I’ll figure out how close together I can process the batches.