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

{
The right issue is harder because it requires provider- and model-specific parameter normalization, preserving compatibility across multiple reasoning pathways, and adding targeted regression coverage. The left issue is comparatively localized to aggregation control flow and cost-calculation tests.
}
https://github.com/berriai/litellm/issues/33401 3:1 https://github.com/berriai/litellm/issues/34730