Creating subsets (or datasets) with 100,000s accession refs is a performance issue for the backend and a UI issue for the frontend: The frontend will time out before the backend responds with results, causing the user to "reupload" the 100,000 refs, using even more backend resources.
Update frontend to send
accessionRef data to backend in batches of 10,000 (or 20,000). It should wait until the server responds and then send the next batch.
select instCode, acceNumb, genus from accession limit 200000;
Also test performance of "Create new version".