Two Assumption. First, that it is the OS. A patch may fix. it.
Assumption 2, that it is the battery:
http://batteryuniversity.com/learn/article/testing_lithium_based_batteries is your starting point. And will quote from there:
"No rapid-test can evaluate all battery symptoms and there are always outliers that defy the test protocol. Correct prediction should be 9 out of 10. QuickSort™ (by Cadex) exceeds this requirement with most Li-ion packs for mobile phones, but this technology only testes single-cell packs up to 1,500mAh. New technologies in development promise to test larger Li-ion packs, but this may extend the test to a few minutes to accommodate low frequency sampling."
https://www.google.ca/url?sa=t&rct=j&q=&esrc=s&source=web&cd=3&ved=0ahUKEwjCl4ze 1KfRAhXI1IMKHQrVAtIQFggpMAI&url=http%3A%…
"QuickSort™ is based on the electrochemical dynamic response of the battery rather than resistance measurement.
Resistance does not provide a reliable indication on capacity fade that occurs with cycling."
All quick health readouts are resistance tests. In all smartphones, Management Controllers just report the resistance, i.e. capacity retention it is at at that moment. Let's say you have a new, yet defective battery. We know they exist. Yet Apple tests all devices, as do the battery manufacturers, right? A quick first resistance testing, it will read 100-105%. Yet it still is defective. It is meant to die in weeks, or months, but, by specs, according to resistance, it is fine. At 86%, it may still meet the 80% above threshold, but still defective and dying.
Cadex testing is down with batteries removed and they are approaching the 30s threshold. But that is another future measurement metric, not yet an industry standard.
So technically Apple staff is correct when referring to its resistance / capacity /'health' measurement standard, one day Fine at 90%, and, 60 days, later, telling Adrian that his battery Died at 20%. Both answers are correct even if days apart. But the dynamic response of the battery was never measured unless you witness that battery removed and placed on special devices... Onboard diagnostics collect MC data, do not stress a battery across a frequency spectrum.
No, these advanced testing techniques are not possible on a production line. 250 000 000 iPhones (and 300 million other brands) at future 30 s Cadex / per battery test means 125 million testing minutes for iPhone intended batteries alone. It can be done post facto to establish the few million defective units, but not possible before, with current technology. The production line, even with 1000 testing kits, or 100 000, would be too slow....
in other words, your intuition is right, and their response is correct against the reference system they use. The issue is inherent to Li Ion technology, limitations and industrial standards.