The current implementation requires double user input (two return presses or two mouse clicks) to invoke an action that is located at the panel header or serves as a button in a modal dialog. The intention was to give a chance to rules to calculate new values to show them to the user before commiting to the action.
But in case where there is no such rule recalculation the need for the double input twice might be even confusing, because the system doesn’t response for the first time (there’s no recalculation and thus there’s no visual feedback).
The system behaviour should be that if there’s no visible data change returned from the server the action should be invoked on the first input.