Sorting LiteLLM GitHub issues by difficulty to implement. Higher = harder.

{
31609 is harder because it requires tracing caller identity through MCP discovery, credential resolution, and subsequent tool invocation while preserving authorization semantics and compatibility across multiple authentication modes. It also carries higher regression risk and needs end-to-end coverage. 32308 is comparatively localized to retry candidate selection and affinity ordering, with focused routing tests likely sufficient.
}
https://github.com/berriai/litellm/issues/31609 3:1 https://github.com/berriai/litellm/issues/32308