Add docs for slow postgres queries

docs/operations.md

@@ -16,7 +16,22 @@ For now, we’re back to manual deployments:
 1. Celebrate! 🎉
 1. If feeds were changed or added since the last deployment, run the **Deploy Feeds** CI job.
 
-## Incident runbook
+## Incident runbooks
 
 Sign in to <https://furrylist.grafana.net> and head to the **Overview**
 dashboard. Look at the colorful metrics and hope for improvement!
+
+### Slow generation due to Postgres query planner misbehaving
+
+If the **CPU usage** and **Feed Generation Avg Duration** are unusually
+high, this can be due to Postgres misbehaving and incorrectly planning
+queries. This can usually be fixed by running the following command in
+psql on production:
+
+```sh
+bff=# analyze verbose candidate_posts;
+INFO:  analyzing "public.candidate_posts"
+INFO:  "candidate_posts": scanned 197354 of 197354 pages, containing 2024025 live rows and 2666 dead rows; 300000 rows in sample, 2024025 estimated total rows
+ANALYZE
+Time: 14113.331 ms (00:14.113)
+```