To extend the closing remarks from a SIGINT perspective, sure some fingerprints are non unique and short lived, have little data. But hang onto it long enough and sure enough some slower data from another band might eventually correlate it with something else.

The last time I looked at this seriously I was trying to find out how much fidelity (if it was possible at all) was necessary to identify someone by their mouse and keyboard input.

It's not just what you do but how you do it.