[device service & experiment service] peerconnection & callback handling #64
Description
Currently many peerconnections remain in the "new" / "failed" state and are not deleted. This causes device-changed callbacks to take longer than expected, since all peerconnections remaining in the status "new" will be checked to see whether a connection may now be established.
Since the services furthermore currently wait for the response of the callbacks this may lead to unexpected request-timeouts.