We can keep the collision list short and, on average, bounded by a small constant because:

A. we can use a hash function that produces uniform hashing reasonably well
B. we can experimentally make the hash table array as large as we want until the number of collisions are minimized, on average, to the point where we would like it to be
C. if the hash table is large enough, only an occasional collision will occur
D. All of the above

D