Sharing a reply from our engineering team for reference –
"There are two major reasons for moving the experience to a UI-blocking one.
First, Enhance models (esp. Denoise) are very GPU-intensive. And increasingly, features that we've been introducing over the past year, such as Lens Blur, plus the new features that we've added to the June release (Reflection Removal, Distracting People Removal) are also very GPU-intensive. They all require running ML models, and these models are not