I definitely understand them not wanting to maintain two parallel settings here. I think this failure sits squarely on the project manager who decided to couple

I definitely understand them not wanting to maintain two parallel settings here. I think this failure sits squarely on the project manager who decided to couple the new behavior (collections instead of frecency) together with a new ui. I'd bet dollars to donuts that they didn't test the changes independently, and they're probably tightly linked in the code. I imagine collections were the driving force here. If the new behavior (collections) were shipped separately, I imagine they would have certainly migrated installations to have a default `recents` collection that users could then easily delete (or migrate and maintain manually)! What a nice situation that would have been as everyone would have gotten exactly what they wanted. If the new UI were tested separately, I can't imagine it would have gotten positive feedback. But maybe I'm just an old grump and most people like it better? It seems like shipping the UI change as a separate feature would have easily identified that the grid view was something a lot of people liked.