When Checkpoints Turn Toxic: The LangGraph Flaw Behind a Misleading Signal Headline

A deserialization bug in an AI checkpointing library shows how persistence features can become execution risks, while the Signal angle remains unconfirmed and should be treated cautiously.