Retrying proxy connect timeouts but not read timeouts

I’ve got a proxy + upstream setup with three defined upstream
servers. I would like nginx to retry requests when there is an error
or when there is a connection timeout (but not a read timeout).
However, it appears that proxy_next_upstream only has the granularity
of “error” and “timeout”.

Is there any way to have proxy_next_upstream differentiate between
connection timeout and read timeout? If not, is creating a ticket in
the bug tracker the proper way to request such a feature?