The swapping alias set in reindexing is really slow. So is validating the content alias. I assume this is because we have so many shards. I think, also, that Elastica is making some unexpected requests during this stage because if I delete indexes while this is happening I get errors. Unrelated indexes.