I would accept "sample Variations" if I was to accept that HIFIMAN did extensive "matched pair" at quality control! those graphs (if I read them correctly) are very close on left & right, if there were no "matched Pair" QC, then L couldn't match R that closely.
But if HIFIMAN were so particular at QC , then sample variation couldn't happen, not to that degree -
am I missing something?
So first I'd like to mention that the use of in-ear mics with that protocol is not a good way to evaluate L/R matching as a general rule (don't over-analyse broad low Q differences, and at higher frequencies any differences full stop).
That being said the high-Q resonances exhibited by that copy of the Sundara aren't caused by the mics themselves as demonstrated by the HD560S and HD650, and other Hifiman planars I've measured in the same way (including the second copy's right channel). Any which way they're so significant that a simple sweep made them easily audible, including the exact frequency where the drop occurred.
You can observe that the resonance on the right driver is a little sharper than on the left, and that there is another one on the same side at nearly twice the frequency.
What you mention is also something that I wondered. Interestingly Hifiman HPs have two serial numbers, one for each driver. The numbers on all three I received so far were successive.
For some Hifiman HPs Solderdude has measured a few copies and it seems that while channel matching is alright for most of them, there are some differences between the samples. Crinacle's measurements of two copies of the Susvara shows alright channel matching for both of them, but the high-Q resonances are in quite different locations.
https://crinacle.com/graphs/headphones/hifiman-susvara/
Perhaps Hifiman manufactures in batches and manages to have good intra-batch consistency but poorer inter-batch one. IDK.