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

{
The left issue is harder because it spans MCP discovery, Responses API event sequencing, streaming lifecycle, HTTP status handling, and provider-error propagation across multiple layers. The right issue is more localized to preserving cancellation semantics and adding targeted regression coverage, although router fallback behavior requires care.
}
https://github.com/berriai/litellm/issues/32561 3:1 https://github.com/berriai/litellm/issues/35329