Another piece of data:
I've measured both the removed resistors with an insulation tester whose voltage can be varied from 0 to 2500V RMS and the current is measurable 0-100uA.
The "low Z" sample reached 100uA at just 30V, corresponding to its ~300k resistance previously measured at both low voltage and 250/500/1000V.
The "open" sample actually conducts fine at higher voltages. It reached 100uA at 1.5kV.
I also tested one of the unopened plugs (one of the set which all measure ~1M at low voltage) and it reached 100uA at 350V, and was linear below that.
None of the samples exhibit traditional voltage dependent resistor behaviour (i.e. a sudden fall in resistance above a certain voltage threshold) at the voltages tested i.e. below 2.5kV RMS.
So this explains how you can get the silly resistance readings but the spark plugs all work just fine.