Repetition is valuable when each run has a question, not when the ensemble simply hopes familiarity will solve the passage. We isolate the transition that destabilises tempo and change one variable at a time. A complete run can conceal improvement because the same recovery happens in a slightly different place.
I stop repeating when attention has become predictive rather than responsive. Practice should sharpen listening, not automate inattention. I repeat a passage under altered acoustics, seating and rehearsal order before calling a problem stable.
The record shows whether the same transition fails across conditions or whether one room and one sequence produced the impression of recurrence. A later musician can focus practice on a defensible pattern instead of repeating the whole work until familiarity merely conceals the unstable point. During rehearsal I change the question attached to each return: balance, timing, or a colleague’s cue.
The sequence stops when performers begin predicting by rote, because repetition has then ceased to provide new evidence about the passage.