You see this in iPhone apps like Gilt. The user scrolls a view, and a subview apparently "sticks" to one edges as the rest of the scrollView slides underneath. That is, there is a text box (or whatever) in the scrollView, that as the scrollView hits the top of the view, then "sticks" there as the rest of the view continues to slide.
So, there are several issues. First, one can determine via "scrollViewDidScroll:" (during normal scrolling) when the view of interest is passing (or re-appearing). There is a fair amount of granularity here - the differences between delegate calls can be a hundred of points or more. That said, when you see the view approach the top of the scrollView, you turn on a second copy of the view statically displayed under the scrollView top. I have not coded this, but it seems like it will lack a real "stick" look - the view will first disappear then reappear.
Second, if one does a setContentOffset:animated, one does not get the delegate messages (Gilt does not do this). So, how do you get the callbacks in this case? Do you use KVO on "scroll.layer.presentationLayer.bounds" ?
Well, I found one way to do this. When the user scrolls by flicking and dragging, the UIScrollView gives its delegate a "scrollViewDidScroll:" message. You can look then to see if the scroller has moved the content to where you need to take some action.
When "sticking" the view, remove it from the scrollView, and then add it to the scrollView's superview (with an origin of 0,0). When unsticking, do the converse.
If you use the UIScrollView setContentOffset:animated:, it gets trickier. What I did was to subclass UIScrollView, use a flag to specify it was setContentOffset moving the offset, then start a fast running timer to monitor contentOffset.
I put the method that handles the math and sticking/unsticking the child view into this subclass. It looks pretty good.