Sorry, my bad. I should have referenced the Minimoog D, (or any other transistor, or diode reverse bias white noise sourced synths) not the SEM which does not offer white noise.
The same problem was discovered on Behringer's DeepMind 12 in earlier OS version of the firmware. It has been since corrected and now the synth has a much better white noise source, devoid of any discernable audio patterns.
In this day and age of super fast DSPs, I can't understand why generating endlessly changing, and non-repeating random values can't be so hard to program ?