There is a Dunhuang copy of the Da fangbian Fo baoenjing 大方便佛報恩經 (The sutra of requiting kindness) at the National Library of China (shelfmark BD01534) which has a number of interesting character variants. One of them is the character 爾 in the phrase ershi 爾時 (at that time), which is extremely common in Chinese translations of Buddhist sutras (e.g. “At that time the Buddha…”). In Buddhist texts, the character in this phrase is invariably written as 尒, which is thus recognized as an alternate form of the standard character 爾. But manuscript BD01534 at the National Library of China writes this variant in a number of different ways:

Of these, the first one is the “normal” way of writing the character, which one would expect in a Buddhist manuscript. The other two are completely unattested in both the lexicographic tradition and – to my knowledge – in other manuscripts. The third form, which is completely wrong as it has the component 心 at the bottom instead of 小, seems to have been influenced by the character 恭 written 6 characters earlier in the same line:

Nevertheless, these anomalous forms seem to imply that the person was largely unaware of not only how this character was supposed to be written but also what it meant in this context. Obviously, he was fully literate because most of the manuscript is written in a “normal” way, yet some characters are problematic. In this case, the distinctly Buddhist way of writing the character 爾 as 尒 suggests that he was completely unfamiliar with Buddhist literature. In some cases, for example, he did not only “misspell” the character 尒 but replaced it altogether with another one, resulting in the meaningless phrase lingshi 令時:

Once again, this points to a complete unfamiliarity of the scribe with Buddhist sutras, where the phrase ershi 尒時 occurs with considerable frequency. Interestingly, this difficulty of reading correctly the Buddhist form 尒 is also attested in the philological tradition. Most notably, the late 6th-century work Jingdian shiwen 經典釋文 quotes the Xiang’er 相爾 commentary to the Laozi 老子, writing it as Xiangyu 相余. William G. Boltz (1982, p. 95, n. 2) comments on this that “this would suggest that he [i.e. Lu Deming 陸德明, the author of the Jingdian shiwen] had never actually seen the text.” Boltz is right in pointing out that the error was the result of misreading 尔, the other common variant of 爾, although we may go a step further and specify that the variant behind the misreading was probably the form 尒 commonly seen in medieval Buddhist manuscripts. This, in turn, suggests that the error may have been introduced not by Lu Deming himself but occurred in the course of the transmission of the Jingdian shiwen. This is also corroborated by the fact that in contrast with this erroneous case, the title of the Xiang’er commentary appears elsewhere in the Jingdian shiwen correctly on at least two occasions.
But coming back to manuscript BD01534, we can also see a number of other variants that are unusual. One of these is the presence of several examples of the Empress Wu character 圀, normally used during the reign of Wu Zetian 武則天 (690-705) in place of the standard form of 國 (country, state). In terms of its physical appearance, however, the manuscript seems to be from the 9th-10th century, although it is undated and thus we cannot be certain about this. We should also keep in mind that there are known examples of Empress Wu characters being used in later manuscripts because they were copied over from earlier ones (cf. Drège 1984). Zhang Nan 张楠 (1992), for example, describes how this particular character continued to be used in Yunnan long after the time of Empress Wu. Nevertheless, in the majority of cases the rule of being able to date manuscripts to Empress Wu’s reign based on the characters she enforced seems to hold true.
The manuscript also has a number of orthographic inconsistencies, such as the following examples:



These are, however, not unusual and it is quite common in the Dunhuang manuscripts to see alternate forms of the same character, even if the entire scroll was written by the same person. This is simply part of manuscript culture. But there are still some variants that do not make sense. For example, the second character in the phrase qingjing 清淨 (peaceful and quiet) is very strange:

Here the right side component of the second character is 乎, rather than the usual 爭. Once again, this is a wholly unattested variant and must be considered an error, especially since the character 淨 appears elsewhere within the same manuscript in its correct form.
Now what do all these variants mean? Although I cannot be entirely sure but it seems that the manuscript was produced outside of the sutra-copying tradition to which we attribute most of the Dunhuang scrolls. It was probably copied by someone who was not familiar with Buddhist sutras, which is a problematic scenario in the Buddhist community of medieval Dunhuang. All this means that this particular manuscript deserves further study and, until we can account for the reasons behind its irregular features, should be used with caution.
References:
Boltz, William G., “The religious and philosophical significance of the ‘Hsiang Erh’ Lao Tzu in light of the Ma-wang-tui silk manuscripts,” BSOAS 45.1 (1982): 95-117.
Drège, Jean-Pierre, “Les Caractères de l’impératrice Wu Zetian dans les manuscrits de Dunhuang et Turfan,” BEFEO 73 (1984): 339-354.
Zhang Nan 張楠, “Wu Zhou xinzi ‘guo’ zai Yunnan de liuchuan kaoshi” 武周新字“國”在雲南的流傳考釋, Gugong bowuyuan yuankan 故宮博物院院刊 (1992) 3: 60-61.