Indeed, this is one of the option we are evaluating.
Not as plug-and-play as having native ROS-speaking hardware, but the glue between the two protocols could be automagic / generated (e.g., something like ros1_bridge)