This isn't a short coming of DBT - they just have to be used correctly. One way to tackle small sample size is to use ranking tests for a number of cables and assess using something like spearmans rank correlation. Obviously something like CDPs can be told apart every time and no one contests they sound different. One could argue that if large sample sizes are required for cables then the differences must be so small that if even if it exists it is not worth bothering with - when the money spent elsewhere would swamp the alleged difference by orders of magnitude.