There's also the question of ascertainment bias, whether cases were found in the wet market simply because that's where the authorities looked. I assume you've seen those endless debates, which seem basically unresolvable to me given the limited information available.
The infected lab worker causing exactly one massive superspreading event in the wet market without infecting anyone else, making it look exactly like zoonotic spillover, without leaving any other contacts around Wuhan is much less believable to me. That sounds like a thousand to one odds or much worse.
That's indeed unlikely, but it's a strawman. As I've noted repeatedly above, most lineages die out. If the introduction into humans occurred outside the market (naturally or otherwise), then it's quite likely that patient infected others outside the market; those lineages just died out. I remember Trevor Bedford showing genomic evidence of all the failed introductions into Seattle before the spread finally went exponential, early in the pandemic.
Pekar et al. is basically a more sophisticated version of your argument, with numerical models of the phylogenetic and epidemiological dynamics. I don't think it's convincing (since they chose their power law pretty arbitrarily), but at least they acknowledge and purport to model that stochasticity.
Except that SARS-CoV-2's epidemiological dynamics are well-known to be overdispersed? Almost all lineages die out, and a few explode due to repeated super-spreader events. It's therefore unlikely that the first cluster will be discovered at the site of introduction. For example, SARS-CoV-2 was presumably first introduced to other continents at airports and seaports; but that's not where the first clusters were found.