on a 5.6 ES cluster that i don't administrate, i noticed, when a query gives an error HTTP 400 such as
{"error":{"root_cause":[{"type":"action_request_validation_exception","reason":"Validation Failed: 1: template is missing;"}]
that it turns to yellow status, and while it re-assigns shards, there is no way to bulk some documents until it comes to 98.5% of active shards.
this is just correlation (i may make a mistake in my judgement)...
is this possible that coordinator nodes can't manage a mal-formed or not valid query ?