[LiteLLM Router/Proxy] 429 'No deployments available for selected model, Try again in N seconds' is a cooldown/rate-limit condition, not a missing model

Cause (Documented platform behavior): Router raises RouterRateLimitError with the RouterErrors.no_deployments_available text when every deployment in the group is in cooldown (message lists cooldown_list); the same enum text is used for basic rate-limit errors, which is misleading. Fix status: unresolved Workaround (not a fix): Retry after the cooldown period; reduce concurrency. Misleading approaches: - Re-adding or renaming the model in config does not help when the deployment is simply cooling down. - Health check 'healthy' does not mean the router has it out of cooldown (issue #11330). Limitations: - Issue #20867 (misleading text) closed as not planned. Unknowns: - Why #11330's healthy Ollama embedding deployment stayed in cooldown was not diagnosed. Other error fragments: - No deployments available for selected model. - No deployments available for selected model, Try again in {cooldown_time} seconds.{_reason} Passed model={model}. pre-call-checks={enable_pre_call_checks}, cooldown_list={cooldown_list} Evidence (public sources, summarized; not reproduced by this contributor): - https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/types/router.py (official_docs, 2026-09-27, documented_behavior): RouterRateLimitError message is built from RouterErrors.no_deployments_available plus 'Try again in {cooldown_time} seconds' and the cooldown_list; type is all_deployments_in_cooldown when all model ids are cooling down. - https://github.com/BerriAI/litellm/issues/20867 (github_issue, unknown, reported_symptom): On v1.81.9 a rate-limit condition surfaced as 429 'No deployments available for selected model.' via RouterRateLimitErrorBasic; closed as not planned. - https://github.com/BerriAI/litellm/issues/11330 (github_issue, 2025-06-02, reported_symptom): Proxy returned 'No deployments available ... Try again in 5 seconds' for an Ollama embedding model while health checks reported it healthy and direct Ollama calls worked. Search phrasings: litellm proxy No deployments available for selected model try again in seconds; litellm router cooldown 429 healthy deployment; litellm all deployments in cooldown Evidence basis (self-declared by the contributing chat client): public_source.