'What is your screen resolution? What HTML5 features do you support? Do you have this font/plugin?'
Bar giving up all those facilities, the best you can do is ask and hope they comply....
For instance, you can change your user agent and headers to some generic Windows Chrome one but a savvy tracker will fingerprint your TCP connection and realise you are actually on a Mac, and that can be used as a further source of entropy to better identify you.
I've spent more time than is healthy looking into ways to uniquely identify devices on the internet. While the company I was working for was always on the safe side in terms of privacy, the more blackhat methods can be useful in collaborating with law enforcement to deanonimise Tor users (think pedophiles, not drugs)
Tor Browser devs have in turn been really good at hardening fingerprinting attack vectors related to it (and rightly so, browsers don't rape people — people rape people) but unless you're super paranoid it's not a great experience for the average user.
I think a better solution will come (at least in the EU) from the proposal to extend cookie law to any kind of fingerprinting (regardless of storage), which takes use case in account and can be enforced through big fines.
Couldn't this be worked around by reporting random values every time?
It's very hard to get away completely... even with Tor, if you are the only person who uses Tor to visit a certain site you are as trackable as anyone else for that site.
It would be interesting to see some reports on that question. Of course, the Law would never abide such accountability, but one can dream...
I don't know about accountability, but we provided support in a proactive way and never more than we were comfortable with (no carte blanche access to data for LEOs, etc.)
The law even has exemptions for security-related applications — your bank already fingerprints your device and uses it as a first line of protection against card theft.