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

{
The left issue requires redesigning asynchronous persistence semantics across several update paths, including completion guarantees, failure propagation, retry or shutdown handling, and regression coverage. The right issue is more localized to request validation and error handling in the MCP gateway, with narrower integration impact.
}
https://github.com/berriai/litellm/issues/35537 3:1 https://github.com/berriai/litellm/issues/32563