Hi there,
There are two feasible ways a single device could be identified via an IPv6 connection (notice I say device and not user!). If the IPv6 interface was not using privacy extensions and used a standard EUI-64 address, a MAC address could be easily identified as it appeared on different IPv6 prefixes. A device would have to have a very old IPv6 network stack for this to occur as privacy extensions being enabled is the more likely scenario.
The second method of identification, depends on the protocol, in particular I am thinking of HTTP/S and browser fingerprinting, which would help identify a browser as its IPv6 (and also IPv4) address changed. That is a passive method of identification.
It may be possible to actively probe an IPv6 (or IPv4) address and derive an OS fingerprint, although this does not give very high fidelity nor many attributes to unique identify a host. It also requires network security devices along the path to permit an active scan and for the host to stay online long enough for it to complete. This is also the least scalable method of identification!
The only way for a user to be tied to multiple IP addresses would be for the user to go through some authentication process and your server to have visibility of the contents of that traffic flow.
cheers,
Seb.