Depending on timing, `t_write_timeout` was getting stuck while checking the resource health, and the previous request timeout options were making a response to never be sent if that process took too long.