yeah, looks very interesting - you have discovered a funny new effect here.
I haven't looked at your tgd, but it's obvious from what I see that your densitiy fractal has a combination of to high contrast/roughness/Noise Variation values. I am not certain though if the roughness is too high here.
In most cases, leave densitiy and roughness *near* default. So if e,g, roughness is at 1.5, then 2.5 isn't *near* anymore, you will probably like to experiment with 1.4 instead in a first step. You can usually leave contrast at 1, and also usually won't increase the noise variation. You are more likely to have to reduce noise variation.
Cheers,
Frank