A bug was exposed while working on SERVER-53023 (creating the TTL index on the recipient).

When we viewed the logs for a failing test after the TTL index ticket changes were made, we saw that a few tenant migrations were started, and each of these completed as well (we saw a "Starting tenant migration ..." and "Marked.." pair for each MigrationID). However, even after the last migration had completed, we found that the OplogBatcher was still running.

Maybe this is because the TenantOplogBatcher is not being joined in the TenantOplogApplier's destructor, for example.