With 30, that gives him 30!/(30-12)! =41,430,393,164,160,000 total permutations. If I followed your example correctly. Big difference depending on the exact number of words in his case, but still quite unlikely that it will be found this way.
The difference is big because it's like 60% of the original dataset.
I don't really know if it's unlikely, because doing time-related calculations is not my specialty. But, in my opinion it's worth knowing how many total arrangements of the words exist. Obviously the great majority of them isn't a valid seed phrase.
Now I'd be more interested to know if there is an open source tool for brute forcing the permutations? Something similar to the vanity address generator of the old days.