elasticsearch bulk indexing (via invoke) sometimes loops over and over due to re-indexing

[naknomum]

manual indexing sometimes becomes stuck in loop and indexing fails

• sometimes happens (especially on large databases) when invoke app.elasticsearch.now --model CLASS is run
• unsure if problem also exists for other bulk indexing (e.g. via celery)
• manifests as progress-bar being interrupted and Serializing (Bulk) Individual [retry=N]: X% message appearing, and progress bar starts over -- usually with a much smaller set of object
• this behavior continues usually indefinitely and has to be interrupted
• when this happens, all indexing progress is usually not persisted to ES and therefore lost
• interrupting and restarting sometimes just works and indexing is done properly
• is a pain on larger production codex when new deployments happen -- makes re-indexing a hassle