Fuzzyhashing vs Fingerprint.js for user device identification?

Question

User identification

Habr
Fingerpring via fingerprintjs DOCs
git
But this library, service has obvious drawbacks described in the article and SO

The article states that Fuzzy Hash is 94-95%, and on the site fingerprint.js 95-98%.
That is, in theory, FP is better than FH, but as described in the article, uniqueness is uniqueness, with the release of a new version of the OS, plugin, browser, or something else, just like with a SHA hash, even 1 bit will change the entire string, hash. What will give an error, they say the user came from an unfamiliar device.

I suppose that Google also uses identification by device, even if you log into your account from another device but with the same IP, for example, a router, there will be a notification about connecting from a new device.

Who has used FuzzyHashing on a live project?
What package?
What percentage of errors?
Is FingerprintJS better?