Found by the skill/knowledge/workflow concurrency E2E effort (ledger: e2e-concurrency-report.md, item O1). With no active LLM provider and no LLM_PROVIDER_API_KEY, a workflow step run stays RUNNING for over 60s instead of failing fast. Expected: fail fast with a clear error so chains reach FAILED promptly and operators get feedback. Repro: start the h2 stack without a key, POST /api/v1/workflows with a valid NATIVE agent, poll the chain.
Found by the skill/knowledge/workflow concurrency E2E effort (ledger: e2e-concurrency-report.md, item O1). With no active LLM provider and no LLM_PROVIDER_API_KEY, a workflow step run stays RUNNING for over 60s instead of failing fast. Expected: fail fast with a clear error so chains reach FAILED promptly and operators get feedback. Repro: start the h2 stack without a key, POST /api/v1/workflows with a valid NATIVE agent, poll the chain.