First, it really doesn't take major effort to make a viral backbone. You'd like to make a protocol that generates them in high multiplex (thousands, millions, billions) and then selects on that background to find functional ones. The current virus could descend from a recombinant generated with such an approach. It might never have been sequenced or observed directly because it was one of innumerable examples that were competitively cultured.

But, this nature piece is really problematic. When removing likely sequencing errors, the independent "spillover events" appear to fit perfectly into a single phylogeny with each node separated by a single mutation. And the A clade descends cleanly from the B clade. There are not enough mutations between them to support a complex explanation like multiple spillovers. This is linked but not explained properly by the nature piece https://virological.org/t/evidence-against-the-veracity-of-s...