We were allowing the client to build up to `MAX_EDIT_PREDICTION_REJECTIONS_PER_REQUEST`. We'll now attempt to flush the rejections when we reach half max. Release Notes: - N/A
Unified · Split
| @@ -993,7 +993,7 @@ impl Zeta { · 992 unmodified lines | ||
| 993 | 993 | }); |
| 994 | 994 | |
| 995 | 995 | let reached_request_limit = |
| 996 | − self.rejected_predictions.len() >= MAX_EDIT_PREDICTION_REJECTIONS_PER_REQUEST; | |
| 996 | + self.rejected_predictions.len() >= MAX_EDIT_PREDICTION_REJECTIONS_PER_REQUEST / 2; | |
| 997 | 997 | let reject_tx = self.reject_predictions_tx.clone(); |
| 998 | 998 | self.reject_predictions_debounce_task = Some(cx.spawn(async move |_this, cx| { |
| 999 | 999 | const REJECT_REQUEST_DEBOUNCE: Duration = Duration::from_secs(15); |