Techniques are disclosed for avoiding conflicting user actions while the users synchronously participate in collaborative video review based on joint state data for a video collaboration session. User actions may conflict, e.g., when a user submits a video playback instruction to change the current frame of the session while another user is performing a frame-specific action on the current frame. The video collaboration service freezes the current frame in joint state data based on detecting that a frame-specific action is being performed or is likely imminent. Detecting a freeze condition may be implicit or explicit. In order to unfreeze the current frame of the joint state data, no active freeze conditions may be in effect. Further, the freeze condition may be lifted implicitly or explicitly. A visual video freeze indication may be displayed by one or more client applications participating in the session while a freeze condition is active.