inconsistent noise for the two domain of different view of the same object?

[abc2cba]

Wonder3D/train_mvdiffusion_image.py

Line 542 in deeba98

timesteps = torch.randint(0, noise_scheduler.num_train_timesteps, (bsz // cfg.num_views,), device=latents.device).repeat_interleave(cfg.num_views)

The comment above the line said "same noise for different views of the same object", by may not the same for the same object of two domain(color and normal)?
Is this a mistaken or by design?