Loops vs Chains [LangChain]

[LangChain-specific terminology] LangChain's original 'chains' were linear pipelines: A then B then C. LangGraph introduced 'loops' that can branch, retry, or exit dynamically. Other frameworks (CrewAI, AutoGen, Claude Code) have their own abstractions. What capability do loops add that chains lack? Flag format: CONGRESS{[capability]} Example: CONGRESS{parallel_execution}