1) change your possibly useful feature to include the ability to turn it off, modify the icon, allow customization.
2) Demand, outraged, it be removed.
Guess which one prevailed.
You don't know which one prevailed. My guess is that it will be #1.
Google teams generally operate on a very rapid release cycle, many with weekly releases. How do you develop a feature that takes several weeks (or months!) when your team releases weekly? What you don't do is branch the code and work on your branch for a long time, either constantly rebasing or trying to do a big merge at the end. That way lies madness. And bugs. Lots and lots of bugs.
Instead, you flag-protect your feature. As much as possible you put your new code into the regularly-exercised paths, so it gets run by automated tests, manual QA and real users, but without actually activating the new functionality. The part that you can't allow to be run, because your feature isn't ready yet, you protect with a flag that defaults to off. You can easily flip it on for your own development testing. When you get far enough along, you can flip it on for a bunch of internal users ("dogfooders").
When it's finally ready to go, you flip the flag for all users. If there are tens of millions of them, you also flip the flag progressively, first for 1%, then wait to see if anything breaks, then progressively greater percentages until you get to 100.
Now... with that in mind, if your feature provokes howls of outrage, what do you do? It's utterly obvious: you flip the flag back off for everyone, to silence the outrage while you figure out the next step.
In this case, I suspect that the next step will be to add a control to the settings interface. That sounds like an easy task, but don't forget that you have to run a gauntlet of UX people focused on keeping the software as simple and intuitive as possible, and a gauntlet of QA people who point out that every boolean option you add doubles their test burden, not to mention internationalization and host of other things. Adding a switch to an app used by a billion people isn't easy. But my guess is that it might make enough sense to do it, so in a few weeks we'll see the feature come back, turned off by default.
Or maybe they'll decide that the clutter and testing burden are too much and just kill it. Could go either way.
But, whatever, it's blindingly obvious that step #1 is to un-flip the damned flag and stop the whining while you figure out what to do next.