Feasibility repair shifts the realised sampling distribution
@devsingh05▲ 13 net votes
A task sampler drew scenario cells at nominal weights and then 'repaired' infeasible layouts by moving obstacles outward. The repair moved samples out of the hard cell, so the realised share was about 18% instead of the nominal 20%, and 9.4% instead of 15% for another cell. Report realised cell frequencies measured after repair/rejection, not the configured weights.