Comment by yorwba

Hmmm... if you need to go deeper (because 1/4 of all hashes have zero leading zeros and zero trailing zeros), you can generalize this by converting the hash into its run-length encoding to get a sequence of rank functions where finding values with the same rank for all rank functions is equivalent to finding hash collisions. Very nice.