I've seen anti-fingerprinting plugins and browser configs, etc, but they all seem to make you more interesting for not being fingerprintable, or for turning off a lot of features used in fingerprinting.

Is there any reason a browser or plugin couldn't just spoof the data used in fingerprinting to random values? So it emits different values for each page or resource you load?

Random is often going to be indirectly fingerprintable too, since some values are always the same (or the same given each browser version). Some values can't be set to be random or otherwise non-fingerprintable data without affecting the user, like viewport size.

Spoofing 'Newly installed most recent version of Google Chrome on the latest version of Windows 11 in the US' without leaving any holes revealing something fishy is surprisingly hard, but that or 'Safari running on recent-ish iPhone running up to date iOS' would be the ideal target to blend in with the crowd.

You can forcefully disable hardware acceleration, that removes a few bits - though javascript can still guesstimate your CPU and memory.

For a more proper solution, you could try to convince Mozilla that granting every website uncapped multi-threading and direct GPU access is a bit much. If webgl/webgpu truly is a must, then maybe each website permission should start with a single-threaded software-renderer that behaves indistinguishable on all supported platforms. And only after granting the website that extra permissions let it determine that a more powerful GPU is available.

I think this is what canvas blocker does.