Draw small grids with a fixed top-left origin. Call routes as direction plus count, stop at checkpoints, and practice resuming from the current cell. Keep the exercise outside a serious Campaign attempt until every role can finish without inventing vocabulary. Then add a timer and one deliberate interruption. Rehearse the full BOMBANANA recovery: hands off, identify the break, declare stability, restart at “Maze,” and collect current fields. Restart from the module name after an interruption, then re-establish stage and current inputs. Resuming from a remembered answer joins two different moments of bomb state.
Measure packet accuracy before raw speed. Count missing fields, ambiguous positions, premature actions, and failed confirmations; only shorten pauses after the packet is reliable. Rotate BOMBANANA roles so everyone experiences observation, manual search, relay translation, and physical execution. Score the rehearsal by missing fields, ambiguous positions, early movements, and failed confirmations. Those counts identify a fix more clearly than completion time alone. Shared understanding is more useful than one player memorizing a table.