It occured to me that the hysteresis of the pzt might be something I could overcome by dithering the pzt around the values that were restored when we brought the system up.
I started the dither at +/-5000, then dithered at +/-1000, then +/-100, then +/-10, then +/-1, with 20 second sleeps between each change.
I dithered both pitch and yaw.
When the beam from the PSL was restored, the IMC REFL beam was completely on the camera, and the IMC locked (though at low value, WFS are off).
I've uploaded the executable file that will automatically dither the pzt to the OPS wiki, under the page name "pzt dither to recover alignment," and copied the executable into the userapps/release/isi/h1/scripts directory, file name 20160606_pzt_dither.txt.