>>43496884
The test is shit, but useful.
There are an absurd number of methods that can be used to uniquely identify a user out of a large stream of traffic.
Being "unique" among any number of users is perfectly acceptable. Being uniquely identified as any particular user is not.
For example, I could constantly change the fonts in my system, change time zones, useragents, system language and external IP address, disable HTML5 Canvas and WebGL, and I would be as unique as any other user but make it far less likely that any entity would be able to tie my identity to another user.