Potential Stability Issue with phase screens

I'm starting to find some strange issues with adding rows to the infinite Von Karman Phase screens. The variance (r0) doesn't look to be preserved now, I know it was previously. Any ideas?

<img width="960" height="720" alt="Image" src="https://github.com/user-attachments/assets/9f08fed3-4140-4f72-9776-de35e07d6952" />