It's possible that residents could infect the raccoon dogs by early 2020. That's a good point. No earlier samples were available, as far as I know. But the question of which scenario of spillover (raccoon dog to humans or vice-versa) is more likely is subjective, in my opinion.
For SARS-1 in 2002/2003, we also think it's more likely that civets infected humans, because civets were more likely to come to contact with bats in the wild, not humans. Because labs were not actively performing gain-of-function experiments on coronaviruses back then, it's undisputed that SARS-1 came from nature. But with SARS-CoV-2, this is less clear. Still, the overall evidence does point that SARS-CoV-2 ancestors came from nature, which is too extensive to explain here. I recommend this long-form that convinced me about this:
Crits-Christoph et al.'s preprint did not specify the SARS-CoV-2 variant, presumably due to the lack of data. But Gao et al.'s preprint (from which Crits-Christoph et al. got the raw data) did specify that the SARS-CoV-2-positive samples share 99.98% to 99.99% similarity to the original Wuhan variant.
