A client commissioned a contract manufacturer in China to make a device I had been developing. Part of the device has two brand-name 100k 0.1% 0402 resistors used as a voltage divider from a precision 2.500V voltage source down to 1.250V used as a voltage reference. They built 100 prototypes of the boards, but there were about 10 failing boards with the same error in firmware - some failing straight away and others failing after a after a few weeks of testing. Found out the cause of the error: Some of the 100K resistors were out by up to 1%, with one measuring 97k... 3% low! The values were all over the shop.
The manufacturer's excuse was they might have had the temperature profile on their reflow oven set a little too high. I suspect this is complete and utter
and it is more likely the manufacturer just had used cheapo generic resistors from the likes that you would see on eBay, AliExpress or a back alleyway behind the Shenzhen market for "a-dime-a-dozen". Even 1% resistors would not have that amount of variance from the same reel. I have never seen resistors going low with too much heat. Going high with excess heat - maybe - but there were no other issues on these boards that were quite complicated with many parts including BGAs.