Thank you Luiz!!!
Im really happy to recieve your idea and it is exactly the out of the box way of thinking that I am also trying to do every now and then.
This is probably pretty cost effective ecpecially on the CPU!
The only problem I see with this is the collision detection on each unit (which I did not mention in my description 
).
I just had another idea yesterday: I might try to use a spline/bezier curve arrange the pieces equally while I am updating the spline ever x frame ( chip off the last piece and add a point where the “head” is @ ). However I heard once that the Array functions like this cost a lot… however there would only be one slightly heavier calculation instead of 10-15 lerps at the same time…