I don't know if it has a particular name, but it is a common problem: fast average case, slow worst case.
A common example is the quicksort algorithm. It is often considered the fastest sorting algorithm in the general case but it is still O(n²) in the worst case. According to statistics, it should never happen in practice, however, with cleverly ordered data, it can. That's why some libraries prefer to use a slightly slower algorithm (ex: heap sort) but with a O(n.log(n)) guarantee, so that it can't be DDoSed.
Unbalanced binary trees can have the same problem, on random data, they work fine, but improperly used, they degenerate into much slower linked lists. Hash tables can also be exploited if they don't used a cryptographically secure hash function, and these are rarely used because they can be so slow as to negate the advantage of using a hash table in the first place.
A common example is the quicksort algorithm. It is often considered the fastest sorting algorithm in the general case but it is still O(n²) in the worst case. According to statistics, it should never happen in practice, however, with cleverly ordered data, it can. That's why some libraries prefer to use a slightly slower algorithm (ex: heap sort) but with a O(n.log(n)) guarantee, so that it can't be DDoSed.
Unbalanced binary trees can have the same problem, on random data, they work fine, but improperly used, they degenerate into much slower linked lists. Hash tables can also be exploited if they don't used a cryptographically secure hash function, and these are rarely used because they can be so slow as to negate the advantage of using a hash table in the first place.