When making SQLite calls via https://esm.town/v/std/sqlite after a long-running operation (like an AI call that takes 2-3 minutes), the SQLite request fails with:

Error: {"statusCode":500,"error":"Internal Server Error","message":"fetch failed"}

The bug occurs because something times out or disconnects during the long AI processing, and when we then try to call SQLite, the connection/context is stale or dead.

1. Open the val's HTTP endpoint (click the endpoint URL)
2. Configure the test:
   • Set delay to 30-180 seconds (simulates AI call duration)
   • Set iterations to 3+ (to catch intermittent failures)
3. Click "Run Repro Test"
4. Wait for the delay + SQLite operation
5. Check results - look for red "FAIL" boxes with "fetch failed"

In dcm31/github-leads-wizard:

1. User submits company to /api/start
2. createSession() does first SQLite INSERT ✅ (works - val is warm)
3. generatePrompt() calls AI - takes 2-3 minutes
4. createSession() or updateSession() does another SQLite call ❌ FAILS

The key insight: SQLite works before the long operation, fails after.

Error: {"statusCode":500,"error":"Internal Server Error","message":"fetch failed"}
    at createResError (https://esm.town/v/std/sqlite?v=6:81:10)
    at Object.execute (https://esm.town/v/std/sqlite?v=6:47:11)
    at eventLoopTick (ext:core/01_core.js:179:7)
    at async createSession (db.ts:40:3)

• Platform: Val Town
• SQLite Library: https://esm.town/v/std/sqlite
• Delay Library: https://esm.town/v/std/delay
• Original val: dcm31/github-leads-wizard

• main.ts - HTTP endpoint with test UI
• README.md - This file
• repro-guide.md - Guide from repro.fyi on making minimal repros