I bet the robots know exactly how many failures before they need to cycle IPs/cookies/accounts and when the time comes I bet they know how to do it quickly and at low cost.

Humans just have to give up and hope that whatever was behind the captcha wasn't actually important.

Many side projects on Github that aren't updated often seems to know the thresholds for not triggering such methods but as I user I often have to pass those stupid verifications.

You'd think they couldn't possibly mimic human mouse movement too, and yet I find taking a stupidly exaggerated wavy line between squares helps. Any attempt to get through quickly and efficiently seems to be deemed robotic in my experience.

Reminds me of a bug I found at (with) work last week: typing too quickly (sending auto-completion requests too quickly) locks up the entire page, nothing for it but to refresh. (It's intended to cancel previous requests since only the latest one is needed, but obviously not supposed to freeze.)