Write a complete blameless postmortem (about one page) for this incident:
On the evening before the CPS 2231 midterm, StudyBuddy's board and Buddy became extremely slow from 20:05 and returned errors from 20:20. Traffic was 4× a normal evening. The database connection pool (20 connections) was exhausted: a new "hardest topics" statistics query, released that afternoon without load testing, took 3 seconds per call and was executed on every board page view. At 20:30, the AI provider also started rejecting requests because the monthly spending limit on the API key was reached (the limit had been set in September for lower usage). Students reported the problem on WeChat at 20:12. Omar was reachable only at 20:50 (no on-call plan). He disabled the statistics feature flag at 21:05 (board recovered at 21:07) and raised the AI spending limit at 21:15 (Buddy recovered at 21:16).
Include: summary and impact, timeline, root causes and contributing factors (at least five), what went well, what went badly, action items (at least six, with owner and due date), and lessons linked to this course (SLOs, canary, load testing, feature flags, monitoring, risk register).