I think the bigger problem is developing an algorithm that reliably solves this problem (and doesn't introduce collateral damage), rather than the computational complexity of the algorithm.