"Select-Object -Unique" and "Group-Object" should not have quadratic runtime
De-duplicating a collection, or aggregating similar items in a collection, can be done in linear time via hash sets or hash tables, respectively.
"Select-Object -unique" and "Group-Object", however, both use O(n^2) algorithms that rely only on equality, never leveraging hashing. It's literally faster (by a lot) to use "Sort-Object -Unique", which additionally does sorting.
Blog post at http://latkin.org/blog/2016/08/02/curious-behavior-when-de-duplicating-a-collection-in-powershell/ has more explanation.
It was suggested by Bruce P on Twitter that I open this bug: https://twitter.com/BrucePayette/status/760532657293832192