Having these two modes be separate introduces a lot of issues due to their subtly different behaviors. We should consider collapsing them into a single mode that targets a window. As part of this initiative, we need to make sure we don't lose any pre-existing capture functionality.
How should the target window be specified? Always target the active window, the window under the cursor, or allow the user a chance to click on a window to target it? I have a light preference for the final option.