[Bug] Claude 3.5 Opus 1M model stops responding after initial task completion

**Bug Description**
I am using the model genuinely , earlier I had raised a question or task to which it sais this its fine but since then the opus 1M model stopped working for other normal tasks

**Environment Info**
- Platform: win32
- Terminal: xterm-256color
- Version: 2.1.274
- Feedback ID: 84f3b26a-d006-443a-a136-eb480f35310f

**Errors**
```json
[{"error":"TelemetrySafeError: VirtualMessageList: itemKeys/messages length desync (keys=512 messages=511 range=[490,512))\n at D7t (B:/~BUN/root/chunk-6wq91xrg.js:37:30963)\n at Lke (B:/~BUN/root/chunk-6wq91xrg.js:37:21857)\n at nu (B:/~BUN/root/chunk-y6sv21ep.js:18:20036)\n at gu (B:/~BUN/root/chunk-y6sv21ep.js:18:38202)\n at Cm (B:/~BUN/root/chunk-y6sv21ep.js:18:84284)\n at Bg (B:/~BUN/root/chunk-y6sv21ep.js:18:83284)\n at Uu (B:/~BUN/root/chunk-y6sv21ep.js:18:83113)\n at ym (B:/~BUN/root/chunk-y6sv21ep.js:18:79637)\n at ct (B:/~BUN/root/chunk-y6sv21ep.js:18:6261)\n at Jn (B:/~BUN/root/chunk-y6sv21ep.js:18:4792)","timestamp":"2026-09-17T09:47:00.389Z"}]
```

No source file or test is identified. Start by reproducing the failure on Claude Code 2.1.274 with the Claude 3.5 Opus 1M model, then inspect the reported VirtualMessageList error in B:/~BUN/root/chunk-6wq91xrg.js and the related stack frames. Done means normal tasks respond after the initial task without the keys/messages desynchronization error.

Written by the indexing model from the issue text.