I took a look of node-curl. It build from its own source code, not depending on node's http.request
Since I haven't used node-curl nor did any load testing on this, I cannot say how good it is under stress. also, I don't know how this node-module handles socket / connection management.
I used http.request before with a lot of load testing and real time massive traffic. So, I can say that one is reliable when you have everything set correctly. Of course, it requires bit more coding.
If you plan to use node-curl or any other modules, I will suggest you to do a lot of load testings, watch the CPU , memory and socket pool, and see if it meets your requirement. I usually do the test by remotely connect to google recaptcha and another node server created by myself in another box - a simply server just have app.get(/*,...) So you have two set of numbers to compare.
Thanks for answer. Right now i'm using Node-Curl. In tests i've been made, node-curl doesn't close connection but http.get closes. Did you ever use node-curl? Do you know any difference between Node-Curl and http.request?