Ask HN: Where do SDK generators break down?

I've been looking the OpenAPI tooling and it seems to be all over the place. Part of it is on the tools, part of it is the nature of generating code, clients/stubs, types, tests and docs for dozens of languages. I also feel like OpenAPI spec isn't equipped for what the tools aim to do, but I lack the experience to say for sure.

To get to the chase, I'm looking for your insights into this space for those with more experience than I. I'm looking for insights into: - Code generator limitations - Client/server stub generator pains - Scenarios where code/spec diverge - Frustrating tooling experiences - And things that work really well in the OpenAPI ecosystem

Thanks