Commands are accumulated each frame via RenderBuffer3D.Add, then executed in insertion order by the active pipeline. The pipeline may re-sort internally if needed for state efficiency (e.g., front-to-back, material batching), but the buffer itself does not impose an order. Uses ArrayPool for the backing store to avoid per-frame heap allocations. The buffer is designed to be cleared and repopulated each frame. RenderBuffer3D.Clear resets the count without deallocating the internal array.