Thanks for your answer. My understanding of your non-desctructive workflow is that you create multiple copies of the original (relatively long) loop, which means even relatively short loops use up the same memory size as the original long loop. Is this correct?
If this is correct, your approach would not work for me: I use very long loops, like one with 512 beats, another one with 1024 beats and yet another one with 2048 beats. At 110bpm these total up to almost 489 seconds (or 8 minutes and 9 seconds). This just fits into the OT’s RAM of 85 MB, which is good for something around 8.5 minutes of stereo material @16bit, which needs to be shared between all flex loaded samples, recording buffers and all pickup loops.
So if I try to do this using your non-desctructive workflow (as far as I have understood or potentially misunderstood it 
), I’d need 3 loops each being 2048 beats long, but only one using its full length, whereas the other uses half of its length (1024 beats), and the final one uses only a quarter of its lengths (512 beats), which would simply not fit into the OT’s RAM, as 3 times 2048 beats@110bpm gives almost 848 seconds or almost 14 minutes.
But again, I may have totally misunderstood your approach.
