The same as they work in any app that is using our API. All activities marked as "async" in the workflow will be executed using the job-executor. So if you perform an action in explorer (e.g.. starting workflow, completing task, …) and the running process encounters an async activity, it will persist the state of the process in the DB and return control to the calling tread. This thread is the request-thread of the container, so the browsers will report back to the user that the action was successful (task completed, process started, ..). After that, the asynchronous processing will be done by the job-excutor in the background.
