This is an appeal to Bayesian statistics. The speaker cannot possibly mean that it is logically impossible for someone to make up whatever-it-is, so they must be meaning that it's extremely improbable. Thus, what they are saying is, effectively, when you are computing the chance of various models given the data, make sure you weight your prior for the "made it up" case appropriately (i.e. very low). Given that, a priori, it's very unlikely that anyone made that thing up, other explanations (even very improbable ones) begin to look correspondingly more favorable.
So it's entirely valid logically (albeit indirect via statistics).
However, although it's potentially valid when doing statistical reasoning, it's not necessarily the case that people intend for you (or themselves) to do it that way. People could be hoping that you will have the same emotional reaction they did: "OMG! No way!!" and then set the a priori probability to zero and thus promote some other perhaps-even-more-wildly-unlikely scenario to be the "best explanation". The correct thing to do may be to recognize that although it's unlikely that someone could come up with such things, everything else is even more unlikely, and thus it-was-made-up is the best explanation (given available evidence).
I'm not sure that a scenario this complicated is well captured by the terms logos, pathos, or ethos. It is like ethos in that arguments from ethos are also arguments about weighting a priori probabilities in a Bayesian framework, but there is no trustworthiness involved here. (Trust is one reason you might adjust your priors.) Although there is logically sound information contained in such an argument, it is usually presented in a very different form than typical appeals to the intellect. Almost no-one is consciously aware of their distribution of prior probabilities, so a more intuitive presentation tends to be more effective (though this says nothing about the validity...).