This is something I haven't prototyped before, so I would appreciate thoughts on how to (not) implement it. The motivation here is simple: when JavaScript queries the results of layout, we want that to be deterministic. If the size of the FrameView/document/etc is not the same on replay as it was when the recording was created, then layout queries are unlikely to be deterministic. A secondary goal is that the window size looks the same on capture and replay, possibly by telling the embedder it needs to resize the view. Showing the intended window size lets the inspector visualize replayed user inputs with overlays. If the window can't actually be resized, we still want layout to use the original size even if not all of the content can be displayed. It would also be nice to make resizing work with capture and replay. Setting the initial size could be treated as a resize command.