And you're confident that everything else bolted together properly (and undamaged) when the rear knuckles where put back in? Back to my earlier comment about how the play was measured. If toe links are bad, or something wasn't put back together properly seems WAY more plausible than bearings being out of spec.
New upper control arms
New lower control arms
New OEM toe links
Knuckles were off a donor car without bearing noise (blown engine though)
Alignment was done
Bearing play is being measured by me grabbing the rim and pulling one side out then pushing it back in. I'd have to do math (and I'm feeling lazy) but the toe on the tire changes about 1/8" or so. I use the term "toe" but the rim will rock back and forth that much in pretty much any direction. The term was just used as a visual aide.
Lifted car up and looked at the suspension while someone moved the rim back and forth. The CV axle moves with the rim, the knuckle and the rest of the suspension does not move at all. Double checked torque on CV axle nut and it was correct. Only option left is bad bearing or bad machine shop.
Regardless of the actual issue, the fix is the same. Get new bearings, rebuild the old set of knuckles (not the donor car knuckles), and try again. I will be using a different machine shop then the one I used last time which should hopefully address a quality of workmanship issue.
Ironically when I ordered the Timken boxed bearings they game with an SKF in the box. Hopefully if the bearings I got were a bad batch then this one will be from a different one. At least I can hope they are. I have verified with SKF's interchange that GRW186 is the correct rear wheel hub bearing.
The wheel bearing is a double angular ball bearing so any play that would allow the wheel to rock back and forth should be measurable as end play. I plan to have the new machine shop put the bearings on a flat supported by the outer race, put a straight edge across it, push down on the inner race, and measure the distance from the outer race to the inner race. They'll then repeat the same process but by supporting the inner race and measure the distance from the straight edge to the outer race. That should give me a total endplay measurement by adding those two figures. I'll then just need to get the actual specs for the bearings from SKF and see if they're within spec.
Then repeat the whole process once they're mounted up in the hubs. They should have near to zero end play once rebuilt in the hub. Runout should also be near zero.