Considering we’re using Tesla & hackney - is there a quick & elegant way of discriminating between {:error, :timeout} that is a TCP connect timeout, versus other reasons of timeout?

What team and myself are currently considering is:

• write own Tesla.Middleware.Timeout with a special :error, :middleware_timeout
• write own hackney adapter for tesla, which would use GitHub - benoitc/hackney: simple HTTP client in Erlang · GitHub when making requests and discriminate between :request_timeout and connect_timeout using this
• fork hackney and modify there and depend on a forked version

None of these options is really appealing at the moment. (the middleware_timeout is easy to pull off, but doesn’t provide full discrimination).