How does hug interact with multiple (gunicorn) workers?

[gerrymanoim]

Apologies if this is obvious, but I just can't seem to find any documentation.

From reading the docs and looking at PR #289 it seems that even though you can have async methods inside hug, all requests will be processed synchronously. If I understand that correctly, if I have an API endpoint that makes a request to a slow DB, the delay in serving that requests will always hold up subsequent requests.

However, if I use gunicorn (or similar server) in front of hug I can spin up multiple workers. Does that mean that a slow DB request being served by one worker won't impact requests hitting another worker? (Maybe depending on the worker type?) That's my reading of the gunicorn docs, but my tests were not conclusive.

In general, are there guidelines for writing performant hug APIs where particular endpoints are slow (external API requests or pyodbc queries)?