No idea, but just to speculate—the court’s goal isn’t actually to scare OpenAI’s users or harm their business, right? It is to collect evidence. Maybe they just figured they don’t need to dip into that pool to get enough evidence.
Who knows, it's probably the judge's twisted idea of "that'd be too far", as if cancelling basic privacy expectations of all users everywhere wouldn't be.