Neat. It took me a while to realise what was going on: the post on Lemmy and the blogpost are two separate entities. The Lemmy post is a link to the blogpost, and the blogpost uses the post_id to fetch the comments (so I guess this means you have to make the blogpost, make the Lemmy post, and then go back and edit the blogpost with the correct id?)

The script is inspectable on the blog - I can see it does:
const url = 'https://lemmy.ml/api/v3/comment/listpost_id=21617067&limit=100&max_depth=8&sort=Top&type_=All';

So I suppose there's an inbuilt limit for comment depth and number of replies, but if you start down the road of working on that, you'll eventually find that you've re-invented a front-end, and there's no end to it.

What the duckquill guys are doing is a bit fudgy, in that they're getting another website to do the federation legwork for them, but the results are pleasing enough.