Those videos of how the 500mm f/4L lenses are made are quite interesting because they show that there are almost two dozen unique steps in the manufacturing process:
1. Material blending
3. Melted glass allows to cool naturally
4. Cut the glass into pieces
10. Shape the glass into sheets
11. Shaping and pressing process
12. Grinning processes
13. Heating the glass and form its shape by pressing (by hands or by automatic machines)
15. Further polishing
16. Rough grinding that produces that curved surface of the lens
17. Fine grinding
18. Polishing and surface curvuture adjustment
19. Optical inspection
20. Clean with ultrasonic washing machines
23. The Lens assembly process itself (done by hand for Canon L lenses)
From a QC perspective, the steps outlined in Red above indicate that there is scope for variation due to human involvement. I always think of buying an 'L' lens as a bespoke product and not a commodity-type consumer one. After all, when you're expected to pay a lot of money for an object that is essentially hand-built and uses the best possible components, you have to expect some degree of variation (all manufacturing processes have tolerances - usually picked-up in the variance or standard deviations of the final goods).
Ultimately, both Nikon and Canon employ highly skilled technicians to build and assemble Professional grade lenses, and there has to be variation between these humans. I would bet a lot of money that if you were to test a large sample of say 2,000 L lenses produced over a 90-day period and created an index of the optical performance of each lens, then plotted the frequency distribution of those L lenses, that the results might look pretty normal (bell-shaped distribution curve), but then take the same data and plot their scores (or their deviations from the mean) chronologically (as a times series plot) or by Technician you would be very surprised indeed. You may find that Technician B on average produces L Lenses with less variation than Technician D or E, or you might find with shift-work that day-shift batches are better/worse than night-shift workers etc. Or even that Assembly line 2 is better than Assembly line 1 or that for Technician A there was a 2-week spike up in August in the variance of his output - that mysteriously dropped when he returned from vacation!
Bloggers on CR often use automobile analogies, well I did quite a bit of research on BMW Motorsport before I ever bought an M-badge car and one feature of their QC process stood out - every M car produced is test-driven on the world famous Nurburgring Racing Circuit, and it is said that the test drivers can tell which Technician hand-built the engine in every car. Furthermore, they had a Chief Technician in the 1990s, who it was said could simply switch on the engine whilst the car was stationary, walk around it and listen to the sound it produced as well as press his ear up to the bodywork (and feel the vibration) of the vehicle and he could then say: "Claus assembled this one!" Urban myth? Perhaps, but it does speak volumes for the fact that even the most expensive items in the world have considerable product-to-product variation.