http://www.wnyc.org/npr_articles/2011/jun/30/how-the-hippies-saved-physics-curious-contributions-to-quantum-theory/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+wnyc_home+(WNYC+New+York+Public+Radio)
Nick, I have assumed both laser beams have identical polarization.

|z1)|z'2') + |z2)|z'1')

If they mix orthogonal polarizations +, - then of course my idea will not work trivially. I have only been considering back-to-back momenta - ignoring the polarizations.

For example, |z1+)|z'2'-) + |z2+)|z'1'-) + |z1-)|z'2+') + |z2-)|z'1'+) might not work?

I am too lazy to do the algebra right now.

The point is that if (+|-) = 0 makes all the cross terms on the untraced half of the pair vanish then there is no nonlocal entanglement signal.

Only if there is a local interference pattern at the receiver without the coincidence circuit do we have a nonlocal signal. We can wash out the local fringes with the Heisenberg microscope on the traced transmitter end introducing a set of random phases independent of the orthogonality or non-orthogonality of the subsystem states.