We're seeing ~10% of current requests being rejected by poolcounter: https://grafana-rw.wikimedia.org/d/qrOStmdGk/elasticsearch-pool-counters?editPanel=2&viewPanel=2&orgId=1&from=1623006099108&to=1623085934837
This is almost certainly due to the ~2x expected `eqiad.full_text` query volume and ~30x expected `eqiad.entity_full_text` query volume: https://grafana.wikimedia.org/d/000000455/elasticsearch-percentiles?viewPanel=47&orgId=1&from=1623010929480&to=1623086544830
We saw an initial spike of 2x `eqiad.full_text` query volume without the corresponding increase in `eqiad.entity_full_text`. During that spike there were [almost] no pool counter rejections, so it seems that the `eqiad.entity_full_text` going up ~30x in the subsequent spike is likely what tipped us over the edge of our capacity. See https://grafana.wikimedia.org/d/000000455/elasticsearch-percentiles?viewPanel=47&orgId=1&from=1623011239023&to=1623033287614 / https://grafana-rw.wikimedia.org/d/qrOStmdGk/elasticsearch-pool-counters?viewPanel=2&orgId=1&from=1623011239023&to=1623033287614 for just that initial spike