See, there's the problem, right there. You could meet every point of that patent on a touchscreen phone using an image of a latch with "Slide to Unlock" written below it. But some legal pedant would still say the idea behind this, when combined with the concept of touch-style drag and drop, which I personally used in 1996 and is a simple extrapolation of the mouse interface which was designed before I was alive, is a new and novel concept.
No, what it comes down to is that you don't understand the difference between "novel" and "obvious". A patent claim is "novel" if it hasn't ever been done before: if I claim a tuna fish and peanut butter sandwich, for example, no one may have ever done that before (with good reason). Therefore, it's "novel". However, if one or more pieces of prior art teach everything in the patent claim and could be easily combined, then it's "obvious": peanut butter sandwiches are known; tuna fish sandwiches are known; it would be trivial to slap the two together. Therefore, even if the pb&t sandwich is novel, it is nonetheless obvious. Clear?
So, let's go back to the Apple slide-to-unlock claim. A physical latch does not, on its own, teach everything in the claim. Therefore, the physical latch does not show that the claim is not novel - or rather, with regard to the physical latch, Apple's patent is novel.
Is it obvious? Maybe... But we need some more prior art that describes the other features in the claim. Otherwise, all you've got is a tuna fish sandwich and no peanut butter, and you can't possibly prove that a pb&t sandwich is obvious.
Hence, millions of patents that basically read "[Something people have been doing for some period of time between a generation and the beginning of recorded history] on a computer/the internet/a tablet/a touchscreen
Nope, there are no patents that read that, and why should now be clear to you: "[Something people have been doing for some period of time between a generation and the beginning of recorded history]" is known; computers/the internet/tablets/touchscreens are known; if it's trivial to do one on the other, then the resulting combination is obvious.
But notice how you hedge - "millions of patents that basically read..." If you have to rewrite the patent claim to say something it's not saying or cut out various features, then you haven't done anything other than prove that your version is obvious. That's as legally and logically sound as saying that a defendant is "basically guilty" so you don't really need evidence.