A literal translation can be grammatical and still miss the decision
Some transferred phrases are simply incorrect. More difficult cases are understandable English that carries the wrong register, emphasis or habitual collocation. The learner may receive little feedback because the sentence is not broken enough to stop communication.
That is why a useful Russian-speaker set should describe the decision, not merely publish a blacklist of mistakes. What does English normally choose here, and what changes when another Frame is used?
Contrast beats a rule that says “do not translate”
Telling learners to “think in English” is not an instruction they can execute on demand. Showing two plausible options and the boundary between them is actionable. The learner can inspect examples, make a Choice, receive feedback and then produce a new sentence.
Over time the English Frame becomes a direct response to the Move instead of a correction applied after a Russian-shaped draft.
Transfer problems are ideal candidates for focused sets
Recurring patterns can be grouped by function: requests, aspect and time, articles, word order, discussion language, register or common lexical frames. The set should prioritise high-value decisions that repeatedly appear in real communication.
The aim is not to eliminate the first language from learning. Russian can explain meaning and sharpen contrasts. The aim is to stop requiring Russian as the hidden template for every English sentence.