> The only thing I find somewhat frustrating about this is that lab leak is nearly impossible to disprove.
A population of animals infected with virus genetically ancestral to the earliest human infections would absolutely disprove any theories of genetic engineering. For example, the discovery of BANAL-20 has pretty strongly excluded any theories that SARS-CoV-2's spike evolved in hACE2 mice (but not that the spike evolved naturally and was used in a laboratory recombinant, as in the unfunded DEFUSE proposal). Such a population wouldn't absolutely disprove theories of research-related spillover without lab manipulation of the virus (e.g., a researcher sampling bat caves gets infected in the field). It would significantly weaken those though, especially if that population was far from any known pre-pandemic sampling trips.
In the previous two coronavirus spillovers (SARS-1, MERS), animals (palm civets and raccoon dogs, camels) infected with virus near-identical to the human virus were discovered within about a year. For MERS we also see genomic evidence of multiple spillovers, separated by an evolutionary distance that makes cryptic zoonotic spread essentially certain. For SARS-CoV-2, we're still waiting. That was the goal of the original pangolin papers; but that turned out to be multiple reports all derived from the same batch of smuggled pangolins, making it much more likely that humans infected those pangolins in the same way that many humans infected their housecats.
The absence of that evidence doesn't mean SARS-CoV-2 is certainly of unnatural origin; the zoonotic path for some unquestionably natural viruses (e.g. Ebola) is still unknown. It's not what anyone expected, though.
> Is there anything we can do about wet markets?
Reckless agricultural practices need to be stopped. The West is guilty too; there's less taste for exotic mammals here, but routine use of antibiotics in healthy animals may be remembered as a crime against humanity. That's a lot harder than stopping a tiny subset of government-funded research with no clear benefits, though. So it seems reasonable to me to work first on that much more achievable goal.