Previous incidents

The issue has been identified and resolved. Some servers were running out of RAM at the same time as a large number of big queries came in. We've updated the configuration so that this is less likely to happen in the future.