The ElasticSearch cluster is a bit weak given it's required workload, and sometimes it will not have enough memory for its indexes. Later versions of ElasticSearch are kind enough to report this fact, rather than crashing like the old versions did. ActiveData already notices this state and issues a `clear_cache` to solve the problem, but AD does not re-try the query. AD should retry the query, and let the client wait longer, rather then returning an error code.
ahal's Test Informant is negatively impacted by this problem