There is little information on Tonkawa, a language once spoken in southern Texas. The earliest data is a word list from 1829, but nothing else was recorded until 1876. The odd thing is that the words recorded from 1829 are either identical to the later ones or, very often, totally different; in the latter case, they are usually transparently derived from different words, usually verbal forms or compounds. No such rapid vocabulary turnover is observed in later work; between 1876 and 1928, nothing much changed. Why would something like this happen?
Apparently, the explanation is remarkably simple, if tragic. The Tonkawa, like many Native American and Australian groups, had a strong taboo against mentioning the name of a dead person, and, when a dead person's name resembled a word in common use, would replace that word. (To reduce this problem, they often gave their children Comanche names.) And, on October 24, 1862, 167 Tonkawas were massacred by an alliance of enemy tribes near Anadarko, Oklahoma. (Given the circumstances of the time, I wouldn't be surprised if this wasn't the only massacre between 1829 and 1876 either.) One side-effect was the sort of rapid vocabulary change that should normally be just about impossible. This is one reason why you don't want to rely on glottochronology too much.
(For more details, see Ives Goddard in ed. Campbell and Mithun, The Languages of Native America, University of Texas: Austin 1979, pp. 358-363.)