What convinced me that the state vectors really don't collapse until observed is the experiment with a laser pointer and three polarized sun-glass lenses.
Point a laser at a piece of polarized stuff oriented up. This will cut the intensity of the beam significantly by filtering out everything that isn't polarized up.
Downstream of that put another polarized piece oriented sideways (90 degrees) from the first piece. Given a sufficiently good quality of polarization this will pretty much totally stop the beam. This is all in accord with everyone's experience.
Now, what happens to the downstream beam (currently totally blocked) if we insert another polarized piece oriented at 45 degrees between the existing two? If the polarized pieces were just acting as filters the intuitive result would be exactly nothing.
This is not what happens. What does happen is that where there was darkness downstream there is now a pretty good beam. What happens at each polarization stage isn't a simple filtering, it is an honest to God observation, and it really does alter the polarization of the incoming photon.
I, for one, welcome our new Observing Overlords /.
This is pretty cool stuff!
What too many authors bent on teleological interpretations or metaphysics don't understand (and, I'm afraid, too many adherents of the Copenhagen School did not appreciate) is that the experimental design is itself the most important part of the observation, as you imply. In graduate school our prof in the second Quantum Mechanics course passed out a mimeographed essay (this was 1979) entitled: Congratulations! You've Created an Apparatus That Might Kill a Cat: It's Really Not That Interesting. It was a terse rant against the Tao of Physics and other alleged profundities then plaguing the scene. I have been hoping it would surface again on the Internet, but have not yet seen it. A pity.