That's just not how the math works. The reality is that the choices of individual consumers is indistinguishable from statistical noise in the data. A single error in a spreadsheet formula has more impact than an individual's choices. No, as systems get bigger and more complex it does not mean that the avenues for individual choices become more numerous. It means that individual choices get averaged out. Again, more consumption patterns are changed by people dying then people making private ethical choices.
Buying one chocolate bar statistically has a direct impact on the extent of child labor in the world. There's no way to know which specific chocolate bars may trigger small tipping points in companies along the line choosing to adjust their production up or down slightly, but every single chocolate bar purchase plays a role in maintaining the child labor used by the industry.
No. It doesn't not. This is a ridiculous claim. The very first thing that will happen if sales lag because of your choices is that retailers will shift to places where the demand is more reliable. The bars easily transported and they are shelf stable. They can be moved to wherever demand for chocolate is increasing. The larger the system, the more easily it can absorb tiny temporary blips in consumption patterns. This is why what is required is organization, because you have to be capable of bringing the movement into a sufficient number of markets such that moving the product around would not avoid the hit to sales.
The impact of buying a chocolate bar on child labor is incredibly small
Again, it's literally non-existent. Children don't make chocolate bars. They harvest and process cacao. Their product is fungible. If the bars sell less, they just move the product of the child labor down market. But even that scenario is ridiculous because your personal choices are indistinguishable from random noise in consumption data. The larger the market, the more organized you need to be to cross the threshold of random noise.
there is no material basis to say that the end products are disconnected from the workers who made them
No one is claiming this. My claim is that this connection is insufficient to be used as leverage by simply choosing to not buy the chocolate bars in private. The labor was already paid for, the bars were already paid for. Your choices have to rise to the quantity required to be noticed statistically in a data set of large scale consumption, and unless you're buying 100x more than a normal person, your choices won't be noticed in the data.
The only reason to double down on claiming that is out of moral discomfort
No one is making the claim in the first place let alone doubling down on it