If the Zorro.exe process memory footprint does not decrease, this does not imply a memory leak. It's possible the memory is being recycled for the next dataset to increase speed. Heap allocations can be slow.
Well, it isn't dataNew, but it's very weird. I could copy the failing script to a new file and it would run without problems. Same thing happened with another script. No idea if it's filename length or what. Perhaps a Wine glitch. No time now to copy stuff over to a real Win & try there... For now close this pending further debugging when time allows.