I have no objection in principle to eating roadkill. I just haven't happened upon any roadkill that I would want, nor have the time to eat. But if I saw, say, a dead deer in the middle of the wild, I wouldn't see any problem with taking the time to prepare it and eating it.
The only reason I can think of, from a purely consequentialist point of view, is if you think someone else, who readily buys meat, is likely to eat it. Suppose somebody who readily buys meat at the grocery store, had they happened upon the dead deer, would have picked up the deer and prepared and eaten it. However, you got to it before they did. In that case, the meat-buyer would likely buy more meat than they would have, had they taken the deer. Thus, you indirectly caused more meat to be bought in the grocery store. If you're a consequentialist, like I am, then that's just as bad as buying the meat yourself. In other words, you might be paying the opportunity cost of somebody else buying less meat.
However, assuming that this isn't the case (for example, if you're in the middle of nowhere and you know that it's extremely unlikely for someone else to find the carcass), there is no problem. Unlike with buying meat at the store, eating the random carcass would not contribute demand to any industry of killing animals, and would not cause more to be killed.
The fear of getting used to meat does not sound compelling to me. First of all, will we invoke that when lab-grown meat comes to the market? And secondly, it does not offer a reason for why the act is bad in and of itself, only that it might compel you to do bad things later. (It is of a similar nature to judgments of killing in simulated environments such as video games. Whether they would make you more compelled to kill in real life says nothing about the actual morality of the act of killing in simulations.)