the theory is not valid, because partially matching hash160 are uniformly distributed over the entire key space and over a large range they will give (with 8 matching HEX standard) a uniform distribution of 1/4.3 billion keys. but in practice I have seen partially matching hash160 even after 500,000 hash160 from the previous one. the probability that by jumping over the range after a partial fall you missed the desired hash160 is minimal, but not equal to 0. so either a probabilistic approach or any complications of a full sequential search that reduce its speed.