In Septuagint studies a common caution against appealing to wildly speculative translator exegesis to account for divergences between MT and LXX is the recognition that the translators were working with a text they recognized as authoritative and unique, and so would have been reluctant to deviate much from the Vorlage. This been confirmed to some degree in a few LXX books where research (particularly of the Finnish school) confirms a high degree of fidelity to the Vorlage combined with dynamic equivalency. In these books, many seeming divergences actually fall within the semantic scope of the Hebrew, if they’re not mistakes or derived from a distinct Vorlage. I think caution is in order, though, and I’ll explain why.
E. Y. Kutcher points to an interesting observation regarding texts found at Masada and their relationship to the standardized manuscripts:
It is interesting to realize that the text of Ben Sira underwent many changes resulting from the “corrections” of medieval (and earlier) scribes. . . . But Psalms fared differently. Except for a few cases of defective spellings, that are also common in our mss of the Bible, there is practically no difference between the text discovered as Masada and out Masoretic text. How are we to account for this difference between the transmission of Psalms and of Ben Sira? The answer is simply that Psalms represented a sacred text and therefore the scribes made every effort to copy it faithfully, while Ben Sira was not canonized, and so it was treated less carefully. This is a clear proof of how particular the scribes were not to change anything when copying a Biblical text.
The texts from Masada clearly come from a later date than the translation of the Septuagint, though (the majority of it, anyway), and the Dead Sea Scrolls show a greater degree of variety in the earliest texts. It seems that an ideology developed between the translation of the Septuagint and the standardization of the MT that saw the text itself as intimately associated with the authority of the message it conveyed. That is, while early translators used dynamic equivalents to convey the sense of the Hebrew without necessarily conveying the original word order, syntax, or lexical qualities, as time went on, the word order, syntax, and lexical qualities became equally as important. What the text was saying was not all that had meaning. How the text said it began to have meaning as well.
This belief became increasingly important over the next few centuries. The second century CE Aquila adhered slavishly to the word order and syntax of the Hebrew in his translation into Greek. His text may be thought of as the original interlinear Bible. Later Jewish scholars would institute strict standards of transcription in order to ensure no part of the text was corrupted, such as counting the numbers of letters in a book. Glaring errors were left in the books with annotations in the margins. The text then took on an entirely new persona. Today it is even viewed as inerrant by many.
I would suggest that retrojecting modern concepts of canonicity and inerrancy into our investigations of the translation of the Septuagint is a bit presentistic. Those concepts have their roots in that time period, but in its infancy I think the idea of reverencing the text of the Bible was quite distinct.
 Anneli Aejmelaeus and Bénédicte Lemmelijn have also argued, convincingly in my opinion, that harmonizations are more likely the work of scribes rather than translators, given the far more broad scope of focus in transcription compared to translation.
 E. Y. Kutscher, A History of the Hebrew Language (Leiden: Brill, 1982), 92.