> reCAPTCHA actually enters an endless selection where it's impossible to pass

I believe this is because from the same IP address there are other users who are failing the recaptcha. So in between you starting, and clicking 'submit', there are a bunch of other wrong-answer recaptchas, so, when combined with your correct answer, it looks like you got it correct just by chance, not by being a human.

This might be true for massive sites (eg Google) but smaller sites only if the cdn aggregates traffic to many different sites.

recaptchas logic to check if someone is a bot or not is global and run on googles servers, not site specific.

Any specific suggestions for experiencing the difference? I've used Tor a bit, and can't say I noticed a huge amount of difference between that and browsing on a new mobile device.

If you're adventurous you could log into Google, PayPal and social media sites. Once you get past captcha challenges and manage to access your account, they may helpfully lock your account and never let you use it again, regardless of how much proof you provide for the ownership of the account to their support staff.

I see! That's something I've never done, I've mostly been using it to look at specific sites where I already knew the URL.

I wish a new web would form inside Tor instead so we could all leave the old compromised web behind.