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

{
The right-hand task is harder because it requires tracing model identity through Azure router selection and the spend-logging pipeline, preserving consistency across response metadata, callbacks, and persistence, with broader regression-testing risk. The left-hand task is a narrowly scoped retry-parameter propagation fix with an established pattern to follow.
}
https://github.com/berriai/litellm/issues/27942 4:1 https://github.com/berriai/litellm/issues/27363