Research questionHow can industrial root-cause analysis reconcile conflicting free-text hypotheses without slow, deeply agentic LLM workflows?Industrial incidents can produce multiple textual root-cause hypotheses that conflict or duplicate one another. Turning these signals into a usable consensus is difficult when deeper LLM workflows introduce context limits, hallucinations, and high inference latency.