The argument is about the likelihood that two researchers independently came up with essentially the same phrase. In other words, how likely is it that two researchers who independently thought up the same idea would also independently take to the same phrase? If, out of the possible phrases that Darwin could have come up with, in order to express the same idea as Matthew (1831), the chance that he hit on the same phrase only with the words in it shuffled, was as small as winning the lottery, then it would indeed by highly unlikely.
But before the plagiarism theorists even gets there, however, they reduces the population of possible ways to express the idea of evolution through natural selection by their own premise, that is, they assume that "natural process of selection" (Matthew 1831) and "process of natural selection" (Darwin 1859) were the only grammatically correct permutations of the four words in the phrase, and they argue that this strengthens the case for plagiarism. On the contrary, it truly weakens the plagiarism claim, because Matthew coming up with his one and Darwin with the other of two viable alternatives would be as as likely as throwing a coin two times ending up with the same results.
Of course, they do not mean that. They believes that there is a gazillion of other possible ways to express the idea of evolution through natural selection and ending up with "natural process of selection" and "process of natural selection" is as unlikely as winning a lottery jackpot. But let's not assume that and check it instead.
Both Matthew and Darwin were dealing with plant and animal breeding and the household term among breeders for what they were doing was selection.
I personally think that the likelihood that both independently fell upon selection was very high, but let's admit that alternatives existed, for example choice and culling. With these alternatives, let's assume that Darwin's probability to also choose 'selection' was 1/3 given the prior probability that Matthew had already chosen 'selection' = 1 (see Bayesian probability on prior probabilities). The likelihood that Darwin also chose 'selection' then was not as high as in tossing a coin twice, but it was far higher than winning a lottery.
Now, given the prior that both have independently chosen 'selection,' how likely would it have been that both also independently chose 'natural,' in order to distinguish their idea about selection going on in nature from the process (oops - can hardly think of other terms myself:-) that breeders are employing? I personally think that the likelihood of this would have been very high, because the alternatives sound awful: innate selection, involuntary selection, non-human selection, selection in the wild? Those are poor alternatives, when you want to distinguish the process of selection by humans from that going on in nature. Environmental selection seems to be about the only bearable alternative. So factors up to P = 1/3 times 1/2 = 1/6 for Darwin to come up with "natural" + "selection" rather than any alternative.
Same goes for 'process.' I have already, inadvertently used the term process above. There are some alternatives, admittedly, but by far not as many and as good ones as plagiarism theorists try to make us believe. Just look at some alternatives: course of natural selection, law of natural selection, mechanism of natural selection, working of ..., procedure ..., action ... . I'd say 'law' would have been a good alternative both might have fallen back on. And they did both speak of a law elsewhere. For the benefit f doubt, let's say all six alternatives were viable. That would leave us with P = 1/6 times 1/6 = 1/36 — the probability that both Darwin threw the same pair with two dice given the prior that Matthew had already thrown that pair. Admittedly, that is unlikely, but by far not as unlikely as winning the jackpot in a lottery.
Alternatives that do not start with selection or a synonym of it in the first place are also not many. 'Survival of the fittest' is an obvious one. But for now, I conclude that starting from the presumption of innocence (something any lawyer should do) the restrictions of language are far greater than assumed by plagiarism theorists. The likelihood that both Mathew and Darwin ended up with similar words, given that they had independently arrived at the same idea (presumption of innocence), did not verge on a miracle.
Apart from logic, language skills are another issue here. It's simply not true that 'natural process of selection' and 'process of natural selection' are the only two grammatically correct permutations of the four words into one phrase. For example: "The selection process of natural varieties led to ..." is a perfect sentence containing a permutation that the plagiarism theorists regard as ungrammatical.