[Candidate / unverified] Solution for 65a6b9fd-81c1-4bb5-8361-841bbc68470d: Run claude update (>=2.1.156 on Opus 4.7/4.8), then /rewind or Esc twice to a ch

## Support semantics This is a **candidate / unverified** Solution from the KFA Solution Factory Tier-2 gate. It is source-grounded research material, not an Outcome, and does **not** claim independent reproduction or confirmed field success. ## Cause The API rejects histories where tool_use/tool_result/thinking block sequences are inconsistent; pre-2.1.156 clients could leave unreparable state where /rewind does not help until upgraded. ## Steps 1. Run claude update to at least v2.1.156 if on Opus 4.7/4.8. 2. Run /rewind or press Esc twice to restore a checkpoint before the corrupted turn. 3. Retry the conversation from that checkpoint. 4. If recovery is impossible, /clear starts a fresh session (loses history). ## Expected result Subsequent API turns succeed after history is rewound to a consistent checkpoint; updated client avoids re-corrupting history during normal tool use. ## Misleading approaches - Retrying the same corrupted session without /rewind on a pre-2.1.156 build. ## Limitations - /rewind may not clear corruption on versions before v2.1.156 during normal tool use per docs. - Not reproduced in this factory lane. ## Unknowns - Whether a specific corrupted session can be salvaged without /clear after rewind. ## Evidence claims - (documented_platform_behavior) 400 tool_use/tool_result/thinking mismatch errors mean the block sequence in history is invalid for the API. — https://code.claude.com/docs/en/errors - (documented_platform_behavior) Versions before v2.1.156 can cause this during normal tool use where /rewind does not clear it; claude update is recommended. — https://code.claude.com/docs/en/errors - (documented_workaround) /rewind or Esc twice recovers from corrupted turns when recovery is possible. — https://code.claude.com/docs/en/errors