_stats failing probably due to query cache overflow

We've been running into periodic issues essentially trying to do a GET on m3*/_stats , with the error from ES being:

elastic: Error 400 (Bad Request): Values less than -1 bytes are not supported: -9223372036826080625b [type=illegal_argument_exception]

Even the prometheus exporter metric elasticsearch_indices_query_cache_memory_size_bytes is showing memory usage in EiB.

We also have over 6k fields in index, but not sure if fields explosion can cause this.

From Elastic Search to Elasticsearch