One thing to note is that, whenever a proxy fails, Searxng will be down and return proxy error.

You can set retries to 2 or more, so SearXNG is going to try another proxy.

Currently, there is no rate limiting ( #1273 uses uwsgi, but we have dropped the dependency to it, it should use Redis).

I do have one question: when proxies are implemented, is the host ip included in the rotation?

I'm not sure to understand your question.

Each new request is sent to the different proxies in turn using a cycle.
There is no health check, and the engine status is for the whole engine not for one specific proxy. I have some code to improve that, but I can't promise it is going to be released anytime soon..

Related code:

(Note: pattern is most probably equal to all:// in your case)

I see, thank you! Say I host Searxng on VPS A, and add VPS B as a proxy:

  proxies:
   all://:
      - socks5h://ip-of-VPS-B:8080

Does this mean that the ip of VPS A would not be used to make search requests?

(In my set up, I actually run a socks5 server on VPS A and included that as a "proxy". Was just wondering if that is redundant)

Does this mean that the ip of VPS A would not be used to make search requests?

Yes, with proxies in the network configuration, only the defined proxies are used and nothing else, not even the clear connexion.