Process never finishes in `1.1.405`, `1.1.406`, `1.1.407`, `1.1.408`

**Describe the bug**
We have a large code base that uses a custom library which has a large, auto-generated file of pydantic models. Since `1.1.405` including all following versions we're seeing the issue that `pyright` does not finish. It will consume all of the cores it gets but simply run ad infinitum - our CI pipelines kill the job after an hour. Going back to `1.1.404` finishes in ~5 minutes.

I am willing to debug this but I have no idea where to even start looking and I can't share code.

Start by comparing Pyright 1.1.404 with 1.1.405 on a representative large generated Pydantic-model code base. Since no reproducible code, identify a minimal reduction or profiling result that isolates the regression, then verify that analysis completes and no longer consumes all available cores indefinitely.

Written by the indexing model from the issue text.