At OSLC we haven't dug into how best to handle long running requests, something a server may take a fairly long time to generate a response to. Take for example a submission to a build engine to execute a build, the execution may take hours. Of course, you don't want to initiate a request and have to leave the connection open to wait to receive a response. I'd expect the Automation(Build/Deploy) working group to work on evolving a proposed solution to this. Though there is precedence on solutions for this, as indicated in a blog post by fellow IBMer / OSLCer Bill Higgins.
I'd be interested in hearing any feedback on experience with HTTP REST-based solutions to handling long-running requests.