I also never heard of that 10% rule but it indeed exists. To summarize, the 10% value (or 5%) is a rule of thumbs that mostly applies when you make a sampling **without** replacement (like in a survey for example, where each person is questionned only once). In this case you will create a bias in the estimation of the variability of the measure (see source 3)
There could also be an issue in experiments where observations are not independent. A small sample size will limit the fact that your observations are linked together.
Source : [http://www.tools4dev.org/resources/how-to-choose-a-sample-size/](http://www.tools4dev.org/resources/how-to-choose-a-sample-size/)
[https://www.statisticshowto.com/10-condition/](https://www.statisticshowto.com/10-condition/)
University of Texas explanation : [https://web.ma.utexas.edu/users/mks/M358KInstr/TenPctCond.pdf](https://web.ma.utexas.edu/users/mks/M358KInstr/TenPctCond.pdf)
Latest Answers