Lesson 2: Protecting the service from overload and restarts

Module 9 · Lesson 2 · About 60 minutes · Needs: Lesson 1 of this module (why the server runs as replicas); Module 5 Lesson 1 (the packaged server); Module 6 Lesson 2 (per-call timeouts); no model key

The situation

Monday 5 October 2026, week 5. The first nap study cohort is in the sleep lab, and the notes server goes to production over Streamable HTTP for Priya and Tomas. Nare asks: what happens when a script loops on search_notes, someone pastes a huge log into create_note, a slow search hangs, or a deploy restarts the server mid-call?

Each answer is a limit you can measure. The running example is Nare's search for "nap study", which finds the anchor note lab-sync-2026-09-02.