Hmmh. Good point. One remark I have: That's kind of made for councils. So you get a representative sample of the population. And than you have like 501 individuals to discuss and make policy. I'm not entirely sure, but it feels to me there's a lower boundary with group size. Once you randomly sample just 3 individuals, I'd be surprised it works as I expect you more to end up with randomness (in the decisions as well). Not with representation.
But also doesn't feel like a new problem to me. For example the US Americans sample their juries in a court. On the other hand they don't randomly sample the sheriff. Looks to me someone already put in some thought. And there's extra things. Like extra steps when sampling the jurors. It's not ...here's your jury, off you go... But there's an entire complicated extra process to it. I suppose that might be related to something like the comparatively small group size of such a jury.