We render a grid of small USDZ models in SceneKit, one model per cell, and each model should spin continuously around its own local Y axis (one full turn every ~8 seconds). Simple grids of 7 and 12 cells work. A full 7×7 (49) or 12×12 (144) does not. Today each spinning cell is its own SCNView with rendersContinuously = true and an SCNAction.rotateBy on that cell’s pivot node. On macOS, opening a 7×7 of these views deadlocks SceneKit: layout stalls inside setFrameSize / setBackgroundColor while every render thread waits on the shared irradiance lock. We cap live views at 12 and keep the remaining cells as static bitmap snapshots, mounted in waves of four. Frame rate is already tiered (12 / 24 / 30 fps) and antialiasing is reduced on weaker devices. The same grid runs on macOS, iPad, and iPhone. What is the efficient way to spin every cell of a 7×7 and a 12×12 around its own axis? Specifically: one shared SCNScene (or one MTKView plus SCNRenderer) with many nodes, each rotating locally, instead of one SCNView per cell — and how should we share geometry, avoid per-view render loops, and stay off that irradiance lock during layout?