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

{
The left issue requires coordinating MCP lifecycle instrumentation, synthetic request metadata, alert bookkeeping, and regression coverage across multiple subsystems. The right issue is comparatively localized to provider propagation during vector-store embedding model resolution.
}
https://github.com/berriai/litellm/issues/32774 4:1 https://github.com/berriai/litellm/issues/23980