Alter will say things like this: "the translation of terms on the basis of immediate context--except when it becomes grotesque to do otherwise--is to be resisted as another instance of the heresy of explanation" (page xxxii of The Five Books of Moses).
But it's very easy for a translator to fall into the heresy. Alter himself does. For example, he writes (to explain his translation without confessing or even seeming to recognize his own heresy of explanation):
The other most common designation of the deity is 'elohim, a word that is plural in form (perhaps, though this is far from certain, a plural of "majesty") but that is generally treated grammatically as a singular. "God" is the natural English equivalent, but in some contexts, where the generic character of the name seems prominent, I have rendered it with a lowercase g as "god," and when the name is treated as a plural, especially when the narrative context involves polytheism, I have translated it as "gods." (page xlvi)Notice how Alter gives his audience a different form depending on context, as if the reader in Hebrew (and now in his English) would need the explanation to disambiguate the words. This, as I've suggested, is to steal the wordplay from both the original and the translation. The reader is robbed.
Two other posts today get at this. danielandtonya at Hebrew and Greek Reader have up a fascinating example in "Clean Choices in Translation?." Wayne at Better Bibles Blog piggy backs on Rick's post at The Lamp to argue that there really ought to be "audience-oriented translation."
(Later this week, I'll try to post on a word-play translation of the first section of Numbers 5. There may be an allusion or two then to the "heresy of explanation.")