Sorry to tell you that you're wrong and you didn't make enough research on the subject.
Canvas Fingerprinting is partly based on the GPU being used. You generating a random canvas fingerprint that would show you using a really old GPU would make you look more suspicious than somebody else.
That wikipedia article (
https://en.wikipedia.org/wiki/Canvas_fingerprinting#Uniqueness) shows that Canvas Fingerprinting by itself isn't enough to identify a user.
Then comes audio. Browsers track the amount of audio channels you have on your computer (you can double check on MLA). I highly doubt that having 2 audio channels (most frequent setup) will make you get suspended rapidly. Do you really think that the average user will get suspended because he has two audio channels?
Should I talk about WebGL aswell (also based on GPU) or are the examples above enough?
The browser fingerprint excuse have been used for many years. You could try by yourself and make a in-house solution without relying on MLA and discover that only ONE feature of MLA could be considered crucial. (Not basing this on assumptions, it's been months I'm not using MLA anymore, and I'm not facing any issue with account suspensions (and not talking about only 10 profiles created here)).
Most important rule of thumb is the following: let's imagine you have a gigantic pile of white rice, and in it you have one little piece of brown rice. Which one will stand out the most? The brown rice of course. Well, it's the same with user agent/canvas fingerprinting/webGL/audio channels manipulation. You using an old browser version (that directly points out to MLA btw, as Chrome updates its browser automatically), a unique canvas fingerprint, a unique amount of audio channels aswell as an unusual aspect ratio (who on earth still uses 800x600 monitors nowadays?) will make you way more visible than your neighbor that uses last Chrome version with no altered metrics.