well, each individual browser should not rotate the values to often, every few weeks or months maybe, if at all, or when sessions are cleared. or they could differ per site. but across all browser instances a statistical distribution of values that is similar to the existing distribution would help to ensure that no particular value stands out. given that it is firefox this should also mean that only average firefox values should be used because the browser itself can be detected through checking feature differences which can't be masked as easily.