We still require a label to be supplied in iconOnly mode which makes it work for screen readers, but we didn't build anything in for mouse users to get hint text. Consumers can currently work around this by manually adding the title prop to the items config, but I think we should also just do this automatically since we already have a label.