Planning Before and After Feedback: Allocating Five Complete Planner Calls with Frozen Visual World Models

Planning with a frozen visual world model poses a temporal allocation problem: given $K=5$ complete planner calls, how many calls $r$ should be spent on independent search before acting, and how many should be reserved to replan from fresh observations? We compare precommitted schedules while fixing the model, scorer, call count, and logical function evaluations within each panel. On Wall, a disjoint pilot selects $r=3$; relative to the all-initial-search schedule $r=5$, the registered endpoint changes by $-34.7$ (two-way 95% sensitivity interval $[-45.0,-24.5]$), with a descriptive success change of $+31.6$ percentage points. PointMaze provides a boundary: both registered intervals for selected $r=1$ versus $r=5$ include zero. A matched Wall sham replaces fresh post-execution observations with cached imagined latents while holding later call slots and horizon work fixed; real minus sham is $-23.9$ (two-way 95% sensitivity interval $[-31.9,-16.4]$), supporting fresh-observation conditioning relative to that control. The Wall direction recurs in targeted iCEM-component, DecentCEM-style, and one-checkpoint JEPA-WM adaptations. PushT repeats the fixed $r=4$ versus $r=5$ direction in a prospective train-split policy confirmation and a retrospective validation-split replication, although equal-call restart controls do not establish a task-general feedback mechanism. Planner-call timing is therefore a measurable, task-dependent controller choice under the tested protocols, not a universal replanning rule.