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

{
The second task is harder because it spans provider translation logic, multimodal data-shape compatibility, and cross-provider regression testing, whereas the first is likely a localized proxy metadata regression.
}
https://github.com/berriai/litellm/issues/28232 5:2 https://github.com/berriai/litellm/issues/27460