I really prefer the previous behavior of the snapping. Now it seems to work fine, but you don't get a visual feedback of the snapping until you release the mouse button. I think that it's questionable from the usability point of view. Snapping result should be visible before you release the mouse click, because the user wants to know if the snapping was successful before that, not after. In the previous versions, the shape was atracted to the snapping point when the pointer entered in the snapping threshold, so you could easily see the end result before releasing the mouse button. Now you can't and it feels like you have no control in the process. For instance, if you need to enlarge or shrink a rectangle beyond a guide and you have the snapping on, it is possible that you get the shape snapped to the guide even if you don't want to, and you'll realize that only after you released the mouse button (thinking that the shape wouldn't snap). Besides that, there are still isues when enlarging or shrinking shapes. Some snapping settings will make almost impossible to scale a shape without being atracted to a snapping point (making it go back to the original size or disappearing). I really don't know about the snapping code (in fact, I don't know anything about code) but it "feels" like there should be a "zone" around the mouse pointer to perform the snapping (and the snapping nodes should be aligned to that zone depending on the axis of the transformation to avoid random snapping) and the snapping should be visible before releasing the mouse button. At the end of the movement, right, but not after the mouse button release.
IMO, the current behavior shows some improvements but also some regressions.
Just my 2 cents. Gez.