I think its about the naive "rep_array[hashcode(obj) % bucket_size] = object" which is all too common.
If your rep_array is doing linear probing and/or robin-hood hasing, then incremental hashcodes (such as 1, 2, 3, 4, 5...) is a bad thing. Especially if you're doing both inserts and removals: this sort of incremental pattern would lead to many "runs" where linear probing would perform poorly.
Of course, it isn't very hard to do rep_array[(hashcode(obj) * large_constant_odd_number) % bucket_size] instead and get a good distribution. But the question is whether or not people know about those kinds of steps.
If your rep_array is doing linear probing and/or robin-hood hasing, then incremental hashcodes (such as 1, 2, 3, 4, 5...) is a bad thing. Especially if you're doing both inserts and removals: this sort of incremental pattern would lead to many "runs" where linear probing would perform poorly.
Of course, it isn't very hard to do rep_array[(hashcode(obj) * large_constant_odd_number) % bucket_size] instead and get a good distribution. But the question is whether or not people know about those kinds of steps.