After replacing that washer, the last step in the platform alignment required 1/6 turn on the front screw between the top two sets of measurements:
The last two sets show the sample-to-sample variation with no adjustments, which didn’t amount to much.
Without changing anything else, I then switched from magenta PETG filament to cyan and ran off two more sets of thinwall hollow boxes (in addition to other doodads) over the next two days:
The variation in the center box height from 4.94 mm to 5.00 mm shows that sensing the platform Z-axis position directly on the glass surface actually works the way it should: ±0.03 mm is as good as it gets. Given that my measurement error / eyeballometric averaging on any given box runs around ±0.02 mm, the far corners also seem rock-stable and certainly don’t justify automatic alignment probing and adjustment.
Thinwall hollow boxes make good handouts at 3D printing presentations…