There have been a few videos by hardware reviewers to address the sample size concern. Gamers Nexus tested 3 different CPU models with 20+ CPUs each and found that the biggest variance from lowest to highest performance was under 4% while performance variance in most cases was about 2%
https://www.youtube.com/watch?v=PUeZQ3pky-w
The way CPU manufacturing and binning are done means that cpus in particular will have very minor differences within the same model number.
They also cut the fabric multiple layers high with a giant stamping tool. The more layers of fabric they cut at once the more variance there is between top layer and bottom layer but the cheaper the process is. Higher quality cheap brands will advertise that they only cut x amount vs what the competitors cut so they have less variation.